* [PATCH v4] RISC-V: Support XTheadVector extension
@ 2023-12-29 4:03 Jun Sha (Joshua)
2023-12-29 4:05 ` [PATCH v4] RISC-V: Refactor riscv-vector-builtins-bases.cc Jun Sha (Joshua)
` (8 more replies)
0 siblings, 9 replies; 133+ messages in thread
From: Jun Sha (Joshua) @ 2023-12-29 4:03 UTC (permalink / raw)
To: gcc-patches
Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw,
christoph.muellner, juzhe.zhong, Jun Sha (Joshua),
Jin Ma, Xianmiao Qu
This patch series presents gcc implementation of the XTheadVector
extension [1].
[1] https://github.com/T-head-Semi/thead-extension-spec/
For some vector patterns that cannot be avoided, we use
"!TARGET_XTHEADVECTOR" to disable them in order not to
generate instructions that xtheadvector does not support,
causing 36 changes in vector.md.
For the th. prefix issue, we use current_output_insn and
the ASM_OUTPUT_OPCODE hook instead of directly modifying
patterns in vector.md.
We have run the GCC test suite and can confirm that there
are no regressions.
All the test results can be found in the following links,
Run without xtheadvector:
https://gcc.gnu.org/pipermail/gcc-testresults/2023-December/803686.html
Run with xtheadvector:
https://gcc.gnu.org/pipermail/gcc-testresults/2023-December/803687.html
Furthermore, we have run the tests in
https://github.com/riscv-non-isa/rvv-intrinsic-doc/tree/main/examples,
and all the tests passed.
Co-authored-by: Jin Ma <jinma@linux.alibaba.com>
Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com>
Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu>
RISC-V: Refactor riscv-vector-builtins-bases.cc
RISC-V: Change csr_operand into vector_length_operand for vsetvl patterns
RISC-V: Introduce XTheadVector as a subset of V1.0.0
RISC-V: Adds the prefix "th." for the instructions of XTheadVector
RISC-V: Handle differences between XTheadvector and Vector
RISC-V: Add support for xtheadvector-specific intrinsics
RISC-V: ...
---
gcc/common/config/riscv/riscv-common.cc | 23 +
gcc/config.gcc | 4 +-
gcc/config/riscv/autovec.md | 2 +-
gcc/config/riscv/predicates.md | 8 +-
gcc/config/riscv/riscv-c.cc | 8 +-
gcc/config/riscv/riscv-protos.h | 1 +
gcc/config/riscv/riscv-string.cc | 3 +
gcc/config/riscv/riscv-v.cc | 13 +-
.../riscv/riscv-vector-builtins-bases.cc | 18 +-
.../riscv/riscv-vector-builtins-bases.h | 19 +
.../riscv/riscv-vector-builtins-shapes.cc | 149 +
.../riscv/riscv-vector-builtins-shapes.h | 3 +
.../riscv/riscv-vector-builtins-types.def | 120 +
gcc/config/riscv/riscv-vector-builtins.cc | 315 +-
gcc/config/riscv/riscv-vector-builtins.h | 5 +-
gcc/config/riscv/riscv-vector-switch.def | 150 +-
gcc/config/riscv/riscv.cc | 46 +-
gcc/config/riscv/riscv.h | 4 +
gcc/config/riscv/riscv.opt | 2 +
gcc/config/riscv/riscv_th_vector.h | 49 +
gcc/config/riscv/t-riscv | 16 +
.../riscv/thead-vector-builtins-functions.def | 659 ++++
gcc/config/riscv/thead-vector-builtins.cc | 887 ++++++
gcc/config/riscv/thead-vector-builtins.h | 123 +
gcc/config/riscv/thead-vector.md | 2827 +++++++++++++++++
gcc/config/riscv/vector-iterators.md | 186 +-
gcc/config/riscv/vector.md | 44 +-
.../riscv/predef-__riscv_th_v_intrinsic.c | 11 +
.../gcc.target/riscv/rvv/base/abi-1.c | 2 +-
.../gcc.target/riscv/rvv/base/pragma-1.c | 2 +-
.../gcc.target/riscv/rvv/xtheadvector.c | 13 +
.../riscv/rvv/xtheadvector/prefix.c | 12 +
.../riscv/rvv/xtheadvector/vlb-vsb.c | 68 +
.../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 +
.../riscv/rvv/xtheadvector/vlh-vsh.c | 68 +
.../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 +
.../riscv/rvv/xtheadvector/vlw-vsw.c | 68 +
.../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 +
gcc/testsuite/lib/target-supports.exp | 12 +
39 files changed, 5931 insertions(+), 213 deletions(-)
create mode 100644 gcc/config/riscv/riscv_th_vector.h
create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def
create mode 100644 gcc/config/riscv/thead-vector-builtins.cc
create mode 100644 gcc/config/riscv/thead-vector-builtins.h
create mode 100644 gcc/config/riscv/thead-vector.md
create mode 100644 gcc/testsuite/gcc.target/riscv/predef-__riscv_th_v_intrinsic.c
create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector.c
create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c
create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c
create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c
create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c
create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c
create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c
create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c
^ permalink raw reply [flat|nested] 133+ messages in thread* [PATCH v4] RISC-V: Refactor riscv-vector-builtins-bases.cc 2023-12-29 4:03 [PATCH v4] RISC-V: Support XTheadVector extension Jun Sha (Joshua) @ 2023-12-29 4:05 ` Jun Sha (Joshua) 2023-12-29 4:06 ` [PATCH v4] RISC-V: Change csr_operand into Jun Sha (Joshua) ` (7 subsequent siblings) 8 siblings, 0 replies; 133+ messages in thread From: Jun Sha (Joshua) @ 2023-12-29 4:05 UTC (permalink / raw) To: gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, Jun Sha (Joshua), Jin Ma, Xianmiao Qu This patch moves the definition of the enums lst_type and frm_op_type into riscv-vector-builtins-bases.h and removes the static visibility of fold_fault_load(), so these can be used in other compile units. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (enum lst_type): (enum frm_op_type): move to riscv-vector-builtins-bases.h * config/riscv/riscv-vector-builtins-bases.h (GCC_RISCV_VECTOR_BUILTINS_BASES_H): Add header files. (enum lst_type): move from (enum frm_op_type): riscv-vector-builtins-bases.cc (fold_fault_load): riscv-vector-builtins-bases.cc Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 18 +----------------- .../riscv/riscv-vector-builtins-bases.h | 19 +++++++++++++++++++ 2 files changed, 20 insertions(+), 17 deletions(-) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index d70468542ee..c51affde353 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -48,24 +48,8 @@ using namespace riscv_vector; namespace riscv_vector { -/* Enumerates types of loads/stores operations. - It's only used in here so we don't define it - in riscv-vector-builtins-bases.h. */ -enum lst_type -{ - LST_UNIT_STRIDE, - LST_STRIDED, - LST_INDEXED, -}; - -enum frm_op_type -{ - NO_FRM, - HAS_FRM, -}; - /* Helper function to fold vleff and vlsegff. */ -static gimple * +gimple * fold_fault_load (gimple_folder &f) { /* fold fault_load (const *base, size_t *new_vl, size_t vl) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 131041ea66f..42d0cd17dc1 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -21,8 +21,27 @@ #ifndef GCC_RISCV_VECTOR_BUILTINS_BASES_H #define GCC_RISCV_VECTOR_BUILTINS_BASES_H +#include "gimple.h" +#include "riscv-vector-builtins.h" + namespace riscv_vector { +/* Enumerates types of loads/stores operations. */ +enum lst_type +{ + LST_UNIT_STRIDE, + LST_STRIDED, + LST_INDEXED, +}; + +enum frm_op_type +{ + NO_FRM, + HAS_FRM, +}; + +extern gimple *fold_fault_load (gimple_folder &f); + namespace bases { extern const function_base *const vsetvl; extern const function_base *const vsetvlmax; -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* [PATCH v4] RISC-V: Change csr_operand into 2023-12-29 4:03 [PATCH v4] RISC-V: Support XTheadVector extension Jun Sha (Joshua) 2023-12-29 4:05 ` [PATCH v4] RISC-V: Refactor riscv-vector-builtins-bases.cc Jun Sha (Joshua) @ 2023-12-29 4:06 ` Jun Sha (Joshua) 2023-12-29 4:10 ` [PATCH v4] RISC-V: Change csr_operand into vector_length_operand for vsetvl patterns Jun Sha (Joshua) 2024-01-02 1:35 ` [PATCH v4] RISC-V: Change csr_operand into juzhe.zhong 2023-12-29 4:13 ` [PATCH v4] RISC-V: Introduce XTheadVector as a subset of V1.0.0 Jun Sha (Joshua) ` (6 subsequent siblings) 8 siblings, 2 replies; 133+ messages in thread From: Jun Sha (Joshua) @ 2023-12-29 4:06 UTC (permalink / raw) To: gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, Jun Sha (Joshua), Jin Ma, Xianmiao Qu This patch use vector_length_operand instead of csr_operand for vsetvl patterns, so that changes for vector will not affect scalar patterns using csr_operand in riscv.md. gcc/ChangeLog: * config/riscv/vector.md: Use vector_length_operand for vsetvl patterns. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config/riscv/vector.md | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index f607d768b26..b5a9055cdc4 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -1496,7 +1496,7 @@ (define_insn "@vsetvl<mode>" [(set (match_operand:P 0 "register_operand" "=r") - (unspec:P [(match_operand:P 1 "csr_operand" "rK") + (unspec:P [(match_operand:P 1 "vector_length_operand" "rK") (match_operand 2 "const_int_operand" "i") (match_operand 3 "const_int_operand" "i") (match_operand 4 "const_int_operand" "i") @@ -1542,7 +1542,7 @@ ;; in vsetvl instruction pattern. (define_insn "@vsetvl_discard_result<mode>" [(set (reg:SI VL_REGNUM) - (unspec:SI [(match_operand:P 0 "csr_operand" "rK") + (unspec:SI [(match_operand:P 0 "vector_length_operand" "rK") (match_operand 1 "const_int_operand" "i") (match_operand 2 "const_int_operand" "i")] UNSPEC_VSETVL)) (set (reg:SI VTYPE_REGNUM) @@ -1564,7 +1564,7 @@ ;; such pattern can allow us gain benefits of these optimizations. (define_insn_and_split "@vsetvl<mode>_no_side_effects" [(set (match_operand:P 0 "register_operand" "=r") - (unspec:P [(match_operand:P 1 "csr_operand" "rK") + (unspec:P [(match_operand:P 1 "vector_length_operand" "rK") (match_operand 2 "const_int_operand" "i") (match_operand 3 "const_int_operand" "i") (match_operand 4 "const_int_operand" "i") @@ -1608,7 +1608,7 @@ [(set (match_operand:DI 0 "register_operand") (sign_extend:DI (subreg:SI - (unspec:DI [(match_operand:P 1 "csr_operand") + (unspec:DI [(match_operand:P 1 "vector_length_operand") (match_operand 2 "const_int_operand") (match_operand 3 "const_int_operand") (match_operand 4 "const_int_operand") -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* [PATCH v4] RISC-V: Change csr_operand into vector_length_operand for vsetvl patterns. 2023-12-29 4:06 ` [PATCH v4] RISC-V: Change csr_operand into Jun Sha (Joshua) @ 2023-12-29 4:10 ` Jun Sha (Joshua) 2024-01-02 1:35 ` juzhe.zhong 2024-01-02 1:35 ` [PATCH v4] RISC-V: Change csr_operand into juzhe.zhong 1 sibling, 1 reply; 133+ messages in thread From: Jun Sha (Joshua) @ 2023-12-29 4:10 UTC (permalink / raw) To: gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, Jun Sha (Joshua), Jin Ma, Xianmiao Qu This patch use vector_length_operand instead of csr_operand for vsetvl patterns, so that changes for vector will not affect scalar patterns using csr_operand in riscv.md. gcc/ChangeLog: * config/riscv/vector.md: Use vector_length_operand for vsetvl patterns. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config/riscv/vector.md | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index f607d768b26..b5a9055cdc4 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -1496,7 +1496,7 @@ (define_insn "@vsetvl<mode>" [(set (match_operand:P 0 "register_operand" "=r") - (unspec:P [(match_operand:P 1 "csr_operand" "rK") + (unspec:P [(match_operand:P 1 "vector_length_operand" "rK") (match_operand 2 "const_int_operand" "i") (match_operand 3 "const_int_operand" "i") (match_operand 4 "const_int_operand" "i") @@ -1542,7 +1542,7 @@ ;; in vsetvl instruction pattern. (define_insn "@vsetvl_discard_result<mode>" [(set (reg:SI VL_REGNUM) - (unspec:SI [(match_operand:P 0 "csr_operand" "rK") + (unspec:SI [(match_operand:P 0 "vector_length_operand" "rK") (match_operand 1 "const_int_operand" "i") (match_operand 2 "const_int_operand" "i")] UNSPEC_VSETVL)) (set (reg:SI VTYPE_REGNUM) @@ -1564,7 +1564,7 @@ ;; such pattern can allow us gain benefits of these optimizations. (define_insn_and_split "@vsetvl<mode>_no_side_effects" [(set (match_operand:P 0 "register_operand" "=r") - (unspec:P [(match_operand:P 1 "csr_operand" "rK") + (unspec:P [(match_operand:P 1 "vector_length_operand" "rK") (match_operand 2 "const_int_operand" "i") (match_operand 3 "const_int_operand" "i") (match_operand 4 "const_int_operand" "i") @@ -1608,7 +1608,7 @@ [(set (match_operand:DI 0 "register_operand") (sign_extend:DI (subreg:SI - (unspec:DI [(match_operand:P 1 "csr_operand") + (unspec:DI [(match_operand:P 1 "vector_length_operand") (match_operand 2 "const_int_operand") (match_operand 3 "const_int_operand") (match_operand 4 "const_int_operand") -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: [PATCH v4] RISC-V: Change csr_operand into vector_length_operand for vsetvl patterns. 2023-12-29 4:10 ` [PATCH v4] RISC-V: Change csr_operand into vector_length_operand for vsetvl patterns Jun Sha (Joshua) @ 2024-01-02 1:35 ` juzhe.zhong 2024-01-02 19:50 ` Christoph Müllner 0 siblings, 1 reply; 133+ messages in thread From: juzhe.zhong @ 2024-01-02 1:35 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, cooper.joshua, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 2706 bytes --] LGTM assume you have passed the regression. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2023-12-29 12:10 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v4] RISC-V: Change csr_operand into vector_length_operand for vsetvl patterns. This patch use vector_length_operand instead of csr_operand for vsetvl patterns, so that changes for vector will not affect scalar patterns using csr_operand in riscv.md. gcc/ChangeLog: * config/riscv/vector.md: Use vector_length_operand for vsetvl patterns. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config/riscv/vector.md | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index f607d768b26..b5a9055cdc4 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -1496,7 +1496,7 @@ (define_insn "@vsetvl<mode>" [(set (match_operand:P 0 "register_operand" "=r") - (unspec:P [(match_operand:P 1 "csr_operand" "rK") + (unspec:P [(match_operand:P 1 "vector_length_operand" "rK") (match_operand 2 "const_int_operand" "i") (match_operand 3 "const_int_operand" "i") (match_operand 4 "const_int_operand" "i") @@ -1542,7 +1542,7 @@ ;; in vsetvl instruction pattern. (define_insn "@vsetvl_discard_result<mode>" [(set (reg:SI VL_REGNUM) - (unspec:SI [(match_operand:P 0 "csr_operand" "rK") + (unspec:SI [(match_operand:P 0 "vector_length_operand" "rK") (match_operand 1 "const_int_operand" "i") (match_operand 2 "const_int_operand" "i")] UNSPEC_VSETVL)) (set (reg:SI VTYPE_REGNUM) @@ -1564,7 +1564,7 @@ ;; such pattern can allow us gain benefits of these optimizations. (define_insn_and_split "@vsetvl<mode>_no_side_effects" [(set (match_operand:P 0 "register_operand" "=r") - (unspec:P [(match_operand:P 1 "csr_operand" "rK") + (unspec:P [(match_operand:P 1 "vector_length_operand" "rK") (match_operand 2 "const_int_operand" "i") (match_operand 3 "const_int_operand" "i") (match_operand 4 "const_int_operand" "i") @@ -1608,7 +1608,7 @@ [(set (match_operand:DI 0 "register_operand") (sign_extend:DI (subreg:SI - (unspec:DI [(match_operand:P 1 "csr_operand") + (unspec:DI [(match_operand:P 1 "vector_length_operand") (match_operand 2 "const_int_operand") (match_operand 3 "const_int_operand") (match_operand 4 "const_int_operand") -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: [PATCH v4] RISC-V: Change csr_operand into vector_length_operand for vsetvl patterns. 2024-01-02 1:35 ` juzhe.zhong @ 2024-01-02 19:50 ` Christoph Müllner 0 siblings, 0 replies; 133+ messages in thread From: Christoph Müllner @ 2024-01-02 19:50 UTC (permalink / raw) To: juzhe.zhong Cc: cooper.joshua, gcc-patches, Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, jinma, cooper.qu, Kito Cheng On Tue, Jan 2, 2024 at 2:35 AM juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> wrote: > > LGTM assume you have passed the regression. Committed. I've rebased this patch, validated that there are no regressions with the patch, and reworded the commit message a bit before that. > > > ________________________________ > juzhe.zhong@rivai.ai > > > From: Jun Sha (Joshua) > Date: 2023-12-29 12:10 > To: gcc-patches > CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu > Subject: [PATCH v4] RISC-V: Change csr_operand into vector_length_operand for vsetvl patterns. > This patch use vector_length_operand instead of csr_operand for > vsetvl patterns, so that changes for vector will not affect scalar > patterns using csr_operand in riscv.md. > > gcc/ChangeLog: > > * config/riscv/vector.md: > Use vector_length_operand for vsetvl patterns. > > Co-authored-by: Jin Ma <jinma@linux.alibaba.com> > Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> > Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> > --- > gcc/config/riscv/vector.md | 8 ++++---- > 1 file changed, 4 insertions(+), 4 deletions(-) > > diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md > index f607d768b26..b5a9055cdc4 100644 > --- a/gcc/config/riscv/vector.md > +++ b/gcc/config/riscv/vector.md > @@ -1496,7 +1496,7 @@ > (define_insn "@vsetvl<mode>" > [(set (match_operand:P 0 "register_operand" "=r") > - (unspec:P [(match_operand:P 1 "csr_operand" "rK") > + (unspec:P [(match_operand:P 1 "vector_length_operand" "rK") > (match_operand 2 "const_int_operand" "i") > (match_operand 3 "const_int_operand" "i") > (match_operand 4 "const_int_operand" "i") > @@ -1542,7 +1542,7 @@ > ;; in vsetvl instruction pattern. > (define_insn "@vsetvl_discard_result<mode>" > [(set (reg:SI VL_REGNUM) > - (unspec:SI [(match_operand:P 0 "csr_operand" "rK") > + (unspec:SI [(match_operand:P 0 "vector_length_operand" "rK") > (match_operand 1 "const_int_operand" "i") > (match_operand 2 "const_int_operand" "i")] UNSPEC_VSETVL)) > (set (reg:SI VTYPE_REGNUM) > @@ -1564,7 +1564,7 @@ > ;; such pattern can allow us gain benefits of these optimizations. > (define_insn_and_split "@vsetvl<mode>_no_side_effects" > [(set (match_operand:P 0 "register_operand" "=r") > - (unspec:P [(match_operand:P 1 "csr_operand" "rK") > + (unspec:P [(match_operand:P 1 "vector_length_operand" "rK") > (match_operand 2 "const_int_operand" "i") > (match_operand 3 "const_int_operand" "i") > (match_operand 4 "const_int_operand" "i") > @@ -1608,7 +1608,7 @@ > [(set (match_operand:DI 0 "register_operand") > (sign_extend:DI > (subreg:SI > - (unspec:DI [(match_operand:P 1 "csr_operand") > + (unspec:DI [(match_operand:P 1 "vector_length_operand") > (match_operand 2 "const_int_operand") > (match_operand 3 "const_int_operand") > (match_operand 4 "const_int_operand") > -- > 2.17.1 > > ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: [PATCH v4] RISC-V: Change csr_operand into 2023-12-29 4:06 ` [PATCH v4] RISC-V: Change csr_operand into Jun Sha (Joshua) 2023-12-29 4:10 ` [PATCH v4] RISC-V: Change csr_operand into vector_length_operand for vsetvl patterns Jun Sha (Joshua) @ 2024-01-02 1:35 ` juzhe.zhong 1 sibling, 0 replies; 133+ messages in thread From: juzhe.zhong @ 2024-01-02 1:35 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, cooper.joshua, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 2663 bytes --] LGTM assume you have passed the regression. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2023-12-29 12:06 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v4] RISC-V: Change csr_operand into This patch use vector_length_operand instead of csr_operand for vsetvl patterns, so that changes for vector will not affect scalar patterns using csr_operand in riscv.md. gcc/ChangeLog: * config/riscv/vector.md: Use vector_length_operand for vsetvl patterns. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config/riscv/vector.md | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index f607d768b26..b5a9055cdc4 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -1496,7 +1496,7 @@ (define_insn "@vsetvl<mode>" [(set (match_operand:P 0 "register_operand" "=r") - (unspec:P [(match_operand:P 1 "csr_operand" "rK") + (unspec:P [(match_operand:P 1 "vector_length_operand" "rK") (match_operand 2 "const_int_operand" "i") (match_operand 3 "const_int_operand" "i") (match_operand 4 "const_int_operand" "i") @@ -1542,7 +1542,7 @@ ;; in vsetvl instruction pattern. (define_insn "@vsetvl_discard_result<mode>" [(set (reg:SI VL_REGNUM) - (unspec:SI [(match_operand:P 0 "csr_operand" "rK") + (unspec:SI [(match_operand:P 0 "vector_length_operand" "rK") (match_operand 1 "const_int_operand" "i") (match_operand 2 "const_int_operand" "i")] UNSPEC_VSETVL)) (set (reg:SI VTYPE_REGNUM) @@ -1564,7 +1564,7 @@ ;; such pattern can allow us gain benefits of these optimizations. (define_insn_and_split "@vsetvl<mode>_no_side_effects" [(set (match_operand:P 0 "register_operand" "=r") - (unspec:P [(match_operand:P 1 "csr_operand" "rK") + (unspec:P [(match_operand:P 1 "vector_length_operand" "rK") (match_operand 2 "const_int_operand" "i") (match_operand 3 "const_int_operand" "i") (match_operand 4 "const_int_operand" "i") @@ -1608,7 +1608,7 @@ [(set (match_operand:DI 0 "register_operand") (sign_extend:DI (subreg:SI - (unspec:DI [(match_operand:P 1 "csr_operand") + (unspec:DI [(match_operand:P 1 "vector_length_operand") (match_operand 2 "const_int_operand") (match_operand 3 "const_int_operand") (match_operand 4 "const_int_operand") -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* [PATCH v4] RISC-V: Introduce XTheadVector as a subset of V1.0.0 2023-12-29 4:03 [PATCH v4] RISC-V: Support XTheadVector extension Jun Sha (Joshua) 2023-12-29 4:05 ` [PATCH v4] RISC-V: Refactor riscv-vector-builtins-bases.cc Jun Sha (Joshua) 2023-12-29 4:06 ` [PATCH v4] RISC-V: Change csr_operand into Jun Sha (Joshua) @ 2023-12-29 4:13 ` Jun Sha (Joshua) 2024-01-04 9:28 ` Jun Sha (Joshua) 2023-12-29 4:19 ` [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector Jun Sha (Joshua) ` (5 subsequent siblings) 8 siblings, 1 reply; 133+ messages in thread From: Jun Sha (Joshua) @ 2023-12-29 4:13 UTC (permalink / raw) To: gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, Jun Sha (Joshua), Jin Ma, Xianmiao Qu This patch is to introduce basic XTheadVector support (march string parsing and a test for __riscv_xtheadvector) according to https://github.com/T-head-Semi/thead-extension-spec/ gcc/ChangeLog: * common/config/riscv/riscv-common.cc (riscv_subset_list::parse): Add new vendor extension. * config/riscv/riscv-c.cc (riscv_cpu_cpp_builtins): Add test marco. * config/riscv/riscv.opt: Add new mask. gcc/testsuite/ChangeLog: * gcc.target/riscv/predef-__riscv_th_v_intrinsic.c: New test. * gcc.target/riscv/rvv/xtheadvector.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/common/config/riscv/riscv-common.cc | 23 +++++++++++++++++++ gcc/config/riscv/riscv-c.cc | 8 +++++-- gcc/config/riscv/riscv.opt | 2 ++ .../riscv/predef-__riscv_th_v_intrinsic.c | 11 +++++++++ .../gcc.target/riscv/rvv/xtheadvector.c | 13 +++++++++++ 5 files changed, 55 insertions(+), 2 deletions(-) create mode 100644 gcc/testsuite/gcc.target/riscv/predef-__riscv_th_v_intrinsic.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector.c diff --git a/gcc/common/config/riscv/riscv-common.cc b/gcc/common/config/riscv/riscv-common.cc index f20d179568d..66b20c154a9 100644 --- a/gcc/common/config/riscv/riscv-common.cc +++ b/gcc/common/config/riscv/riscv-common.cc @@ -368,6 +368,7 @@ static const struct riscv_ext_version riscv_ext_version_table[] = {"xtheadmemidx", ISA_SPEC_CLASS_NONE, 1, 0}, {"xtheadmempair", ISA_SPEC_CLASS_NONE, 1, 0}, {"xtheadsync", ISA_SPEC_CLASS_NONE, 1, 0}, + {"xtheadvector", ISA_SPEC_CLASS_NONE, 1, 0}, {"xventanacondops", ISA_SPEC_CLASS_NONE, 1, 0}, @@ -1251,6 +1252,15 @@ riscv_subset_list::check_conflict_ext () if (lookup ("zcmp")) error_at (m_loc, "%<-march=%s%>: zcd conflicts with zcmp", m_arch); } + + if ((lookup ("v") || lookup ("zve32x") + || lookup ("zve64x") || lookup ("zve32f") + || lookup ("zve64f") || lookup ("zve64d") + || lookup ("zvl32b") || lookup ("zvl64b") + || lookup ("zvl128b") || lookup ("zvfh")) + && lookup ("xtheadvector")) + error_at (m_loc, "%<-march=%s%>: xtheadvector conflicts with vector " + "extension or its sub-extensions", m_arch); } /* Parsing function for multi-letter extensions. @@ -1743,6 +1753,19 @@ static const riscv_ext_flag_table_t riscv_ext_flag_table[] = {"xtheadmemidx", &gcc_options::x_riscv_xthead_subext, MASK_XTHEADMEMIDX}, {"xtheadmempair", &gcc_options::x_riscv_xthead_subext, MASK_XTHEADMEMPAIR}, {"xtheadsync", &gcc_options::x_riscv_xthead_subext, MASK_XTHEADSYNC}, + {"xtheadvector", &gcc_options::x_riscv_xthead_subext, MASK_XTHEADVECTOR}, + {"xtheadvector", &gcc_options::x_riscv_vector_elen_flags, MASK_VECTOR_ELEN_32}, + {"xtheadvector", &gcc_options::x_riscv_vector_elen_flags, MASK_VECTOR_ELEN_64}, + {"xtheadvector", &gcc_options::x_riscv_vector_elen_flags, MASK_VECTOR_ELEN_FP_32}, + {"xtheadvector", &gcc_options::x_riscv_vector_elen_flags, MASK_VECTOR_ELEN_FP_64}, + {"xtheadvector", &gcc_options::x_riscv_vector_elen_flags, MASK_VECTOR_ELEN_FP_16}, + {"xtheadvector", &gcc_options::x_riscv_zvl_flags, MASK_ZVL32B}, + {"xtheadvector", &gcc_options::x_riscv_zvl_flags, MASK_ZVL64B}, + {"xtheadvector", &gcc_options::x_riscv_zvl_flags, MASK_ZVL128B}, + {"xtheadvector", &gcc_options::x_riscv_zf_subext, MASK_ZVFHMIN}, + {"xtheadvector", &gcc_options::x_riscv_zf_subext, MASK_ZVFH}, + {"xtheadvector", &gcc_options::x_target_flags, MASK_FULL_V}, + {"xtheadvector", &gcc_options::x_target_flags, MASK_VECTOR}, {"xventanacondops", &gcc_options::x_riscv_xventana_subext, MASK_XVENTANACONDOPS}, diff --git a/gcc/config/riscv/riscv-c.cc b/gcc/config/riscv/riscv-c.cc index d70eb8ed361..d7c63ead147 100644 --- a/gcc/config/riscv/riscv-c.cc +++ b/gcc/config/riscv/riscv-c.cc @@ -138,6 +138,10 @@ riscv_cpu_cpp_builtins (cpp_reader *pfile) riscv_ext_version_value (0, 11)); } + if (TARGET_XTHEADVECTOR) + builtin_define_with_int_value ("__riscv_th_v_intrinsic", + riscv_ext_version_value (0, 11)); + /* Define architecture extension test macros. */ builtin_define_with_int_value ("__riscv_arch_test", 1); @@ -191,8 +195,8 @@ riscv_pragma_intrinsic (cpp_reader *) { if (!TARGET_VECTOR) { - error ("%<#pragma riscv intrinsic%> option %qs needs 'V' extension " - "enabled", + error ("%<#pragma riscv intrinsic%> option %qs needs 'V' or " + "'XTHEADVECTOR' extension enabled", name); return; } diff --git a/gcc/config/riscv/riscv.opt b/gcc/config/riscv/riscv.opt index ede2d655e73..7de5f18e11b 100644 --- a/gcc/config/riscv/riscv.opt +++ b/gcc/config/riscv/riscv.opt @@ -449,6 +449,8 @@ Mask(XTHEADMEMPAIR) Var(riscv_xthead_subext) Mask(XTHEADSYNC) Var(riscv_xthead_subext) +Mask(XTHEADVECTOR) Var(riscv_xthead_subext) + TargetVariable int riscv_xventana_subext diff --git a/gcc/testsuite/gcc.target/riscv/predef-__riscv_th_v_intrinsic.c b/gcc/testsuite/gcc.target/riscv/predef-__riscv_th_v_intrinsic.c new file mode 100644 index 00000000000..1c764241db6 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/predef-__riscv_th_v_intrinsic.c @@ -0,0 +1,11 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64imafdcxtheadvector -mabi=lp64d" } */ + +int main () { + +#if __riscv_th_v_intrinsic != 11000 +#error "__riscv_th_v_intrinsic" +#endif + + return 0; +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector.c new file mode 100644 index 00000000000..d52921e1314 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector.c @@ -0,0 +1,13 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gc_xtheadvector" { target { rv32 } } } */ +/* { dg-options "-march=rv64gc_xtheadvector" { target { rv64 } } } */ + +#ifndef __riscv_xtheadvector +#error "Feature macro not defined" +#endif + +int +foo (int a) +{ + return a; +} \ No newline at end of file -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* [PATCH v4] RISC-V: Introduce XTheadVector as a subset of V1.0.0 2023-12-29 4:13 ` [PATCH v4] RISC-V: Introduce XTheadVector as a subset of V1.0.0 Jun Sha (Joshua) @ 2024-01-04 9:28 ` Jun Sha (Joshua) 0 siblings, 0 replies; 133+ messages in thread From: Jun Sha (Joshua) @ 2024-01-04 9:28 UTC (permalink / raw) To: gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, Jun Sha (Joshua), Jin Ma, Xianmiao Qu This patch is to introduce basic XTheadVector support (march string parsing and a test for __riscv_xtheadvector) according to https://github.com/T-head-Semi/thead-extension-spec/ gcc/ChangeLog: * common/config/riscv/riscv-common.cc (riscv_subset_list::parse): Add new vendor extension. * config/riscv/riscv-c.cc (riscv_cpu_cpp_builtins): Add test marco. * config/riscv/riscv.opt: Add new mask. gcc/testsuite/ChangeLog: * gcc.target/riscv/predef-__riscv_th_v_intrinsic.c: New test. * gcc.target/riscv/rvv/xtheadvector.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/common/config/riscv/riscv-common.cc | 23 +++++++++++++++++++ gcc/config/riscv/riscv-c.cc | 8 +++++-- gcc/config/riscv/riscv.opt | 2 ++ .../riscv/predef-__riscv_th_v_intrinsic.c | 11 +++++++++ .../gcc.target/riscv/rvv/xtheadvector.c | 13 +++++++++++ 5 files changed, 55 insertions(+), 2 deletions(-) create mode 100644 gcc/testsuite/gcc.target/riscv/predef-__riscv_th_v_intrinsic.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector.c diff --git a/gcc/common/config/riscv/riscv-common.cc b/gcc/common/config/riscv/riscv-common.cc index 0301d170a41..449722070d4 100644 --- a/gcc/common/config/riscv/riscv-common.cc +++ b/gcc/common/config/riscv/riscv-common.cc @@ -368,6 +368,7 @@ static const struct riscv_ext_version riscv_ext_version_table[] = {"xtheadmemidx", ISA_SPEC_CLASS_NONE, 1, 0}, {"xtheadmempair", ISA_SPEC_CLASS_NONE, 1, 0}, {"xtheadsync", ISA_SPEC_CLASS_NONE, 1, 0}, + {"xtheadvector", ISA_SPEC_CLASS_NONE, 1, 0}, {"xventanacondops", ISA_SPEC_CLASS_NONE, 1, 0}, @@ -1251,6 +1252,15 @@ riscv_subset_list::check_conflict_ext () if (lookup ("zcmp")) error_at (m_loc, "%<-march=%s%>: zcd conflicts with zcmp", m_arch); } + + if ((lookup ("v") || lookup ("zve32x") + || lookup ("zve64x") || lookup ("zve32f") + || lookup ("zve64f") || lookup ("zve64d") + || lookup ("zvl32b") || lookup ("zvl64b") + || lookup ("zvl128b") || lookup ("zvfh")) + && lookup ("xtheadvector")) + error_at (m_loc, "%<-march=%s%>: xtheadvector conflicts with vector " + "extension or its sub-extensions", m_arch); } /* Parsing function for multi-letter extensions. @@ -1743,6 +1753,19 @@ static const riscv_ext_flag_table_t riscv_ext_flag_table[] = {"xtheadmemidx", &gcc_options::x_riscv_xthead_subext, MASK_XTHEADMEMIDX}, {"xtheadmempair", &gcc_options::x_riscv_xthead_subext, MASK_XTHEADMEMPAIR}, {"xtheadsync", &gcc_options::x_riscv_xthead_subext, MASK_XTHEADSYNC}, + {"xtheadvector", &gcc_options::x_riscv_xthead_subext, MASK_XTHEADVECTOR}, + {"xtheadvector", &gcc_options::x_riscv_vector_elen_flags, MASK_VECTOR_ELEN_32}, + {"xtheadvector", &gcc_options::x_riscv_vector_elen_flags, MASK_VECTOR_ELEN_64}, + {"xtheadvector", &gcc_options::x_riscv_vector_elen_flags, MASK_VECTOR_ELEN_FP_32}, + {"xtheadvector", &gcc_options::x_riscv_vector_elen_flags, MASK_VECTOR_ELEN_FP_64}, + {"xtheadvector", &gcc_options::x_riscv_vector_elen_flags, MASK_VECTOR_ELEN_FP_16}, + {"xtheadvector", &gcc_options::x_riscv_zvl_flags, MASK_ZVL32B}, + {"xtheadvector", &gcc_options::x_riscv_zvl_flags, MASK_ZVL64B}, + {"xtheadvector", &gcc_options::x_riscv_zvl_flags, MASK_ZVL128B}, + {"xtheadvector", &gcc_options::x_riscv_zf_subext, MASK_ZVFHMIN}, + {"xtheadvector", &gcc_options::x_riscv_zf_subext, MASK_ZVFH}, + {"xtheadvector", &gcc_options::x_target_flags, MASK_FULL_V}, + {"xtheadvector", &gcc_options::x_target_flags, MASK_VECTOR}, {"xventanacondops", &gcc_options::x_riscv_xventana_subext, MASK_XVENTANACONDOPS}, diff --git a/gcc/config/riscv/riscv-c.cc b/gcc/config/riscv/riscv-c.cc index ba60cd8b555..422ddc2c308 100644 --- a/gcc/config/riscv/riscv-c.cc +++ b/gcc/config/riscv/riscv-c.cc @@ -142,6 +142,10 @@ riscv_cpu_cpp_builtins (cpp_reader *pfile) riscv_ext_version_value (0, 11)); } + if (TARGET_XTHEADVECTOR) + builtin_define_with_int_value ("__riscv_th_v_intrinsic", + riscv_ext_version_value (0, 11)); + /* Define architecture extension test macros. */ builtin_define_with_int_value ("__riscv_arch_test", 1); @@ -195,8 +199,8 @@ riscv_pragma_intrinsic (cpp_reader *) { if (!TARGET_VECTOR) { - error ("%<#pragma riscv intrinsic%> option %qs needs 'V' extension " - "enabled", + error ("%<#pragma riscv intrinsic%> option %qs needs 'V' or " + "'XTHEADVECTOR' extension enabled", name); return; } diff --git a/gcc/config/riscv/riscv.opt b/gcc/config/riscv/riscv.opt index 44ed6d69da2..bb18a22b693 100644 --- a/gcc/config/riscv/riscv.opt +++ b/gcc/config/riscv/riscv.opt @@ -452,6 +452,8 @@ Mask(XTHEADMEMPAIR) Var(riscv_xthead_subext) Mask(XTHEADSYNC) Var(riscv_xthead_subext) +Mask(XTHEADVECTOR) Var(riscv_xthead_subext) + TargetVariable int riscv_xventana_subext diff --git a/gcc/testsuite/gcc.target/riscv/predef-__riscv_th_v_intrinsic.c b/gcc/testsuite/gcc.target/riscv/predef-__riscv_th_v_intrinsic.c new file mode 100644 index 00000000000..550b9039a06 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/predef-__riscv_th_v_intrinsic.c @@ -0,0 +1,11 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64imafdcxtheadvector -mabi=lp64d" } */ + +int main () { + +#if __riscv_th_v_intrinsic != 11000 +#error "__riscv_th_v_intrinsic" +#endif + + return 0; +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector.c new file mode 100644 index 00000000000..8ad370172e3 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector.c @@ -0,0 +1,13 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gc_xtheadvector" { target { rv32 } } } */ +/* { dg-options "-march=rv64gc_xtheadvector" { target { rv64 } } } */ + +#ifndef __riscv_xtheadvector +#error "Feature macro not defined" +#endif + +int +foo (int a) +{ + return a; +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. 2023-12-29 4:03 [PATCH v4] RISC-V: Support XTheadVector extension Jun Sha (Joshua) ` (2 preceding siblings ...) 2023-12-29 4:13 ` [PATCH v4] RISC-V: Introduce XTheadVector as a subset of V1.0.0 Jun Sha (Joshua) @ 2023-12-29 4:19 ` Jun Sha (Joshua) 2023-12-31 17:43 ` Jeff Law 2024-01-03 6:08 ` Jun Sha (Joshua) 2023-12-29 4:21 ` [PATCH v4] RISC-V: Handle differences between XTheadvector and Vector Jun Sha (Joshua) ` (4 subsequent siblings) 8 siblings, 2 replies; 133+ messages in thread From: Jun Sha (Joshua) @ 2023-12-29 4:19 UTC (permalink / raw) To: gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, Jun Sha (Joshua), Jin Ma, Xianmiao Qu This patch adds th. prefix to all XTheadVector instructions by implementing new assembly output functions. We only check the prefix is 'v', so that no extra attribute is needed. gcc/ChangeLog: * config/riscv/riscv-protos.h (riscv_asm_output_opcode): New function to add assembler insn code prefix/suffix. * config/riscv/riscv.cc (riscv_asm_output_opcode): Likewise. * config/riscv/riscv.h (ASM_OUTPUT_OPCODE): Likewise. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config/riscv/riscv-protos.h | 1 + gcc/config/riscv/riscv.cc | 14 ++++++++++++++ gcc/config/riscv/riscv.h | 4 ++++ .../gcc.target/riscv/rvv/xtheadvector/prefix.c | 12 ++++++++++++ 4 files changed, 31 insertions(+) create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c diff --git a/gcc/config/riscv/riscv-protos.h b/gcc/config/riscv/riscv-protos.h index 31049ef7523..5ea54b45703 100644 --- a/gcc/config/riscv/riscv-protos.h +++ b/gcc/config/riscv/riscv-protos.h @@ -102,6 +102,7 @@ struct riscv_address_info { }; /* Routines implemented in riscv.cc. */ +extern const char *riscv_asm_output_opcode (FILE *asm_out_file, const char *p); extern enum riscv_symbol_type riscv_classify_symbolic_expression (rtx); extern bool riscv_symbolic_constant_p (rtx, enum riscv_symbol_type *); extern int riscv_float_const_rtx_index_for_fli (rtx); diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index 0d1cbc5cb5f..ea1d59d9cf2 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -5636,6 +5636,20 @@ riscv_get_v_regno_alignment (machine_mode mode) return lmul; } +/* Define ASM_OUTPUT_OPCODE to do anything special before + emitting an opcode. */ +const char * +riscv_asm_output_opcode (FILE *asm_out_file, const char *p) +{ + /* We need to add th. prefix to all the xtheadvector + insturctions here.*/ + if (TARGET_XTHEADVECTOR && current_output_insn != NULL_RTX && + p[0] == 'v') + fputs ("th.", asm_out_file); + + return p; +} + /* Implement TARGET_PRINT_OPERAND. The RISCV-specific operand codes are: 'h' Print the high-part relocation associated with OP, after stripping diff --git a/gcc/config/riscv/riscv.h b/gcc/config/riscv/riscv.h index 6df9ec73c5e..c33361a254d 100644 --- a/gcc/config/riscv/riscv.h +++ b/gcc/config/riscv/riscv.h @@ -826,6 +826,10 @@ extern enum riscv_cc get_riscv_cc (const rtx use); asm_fprintf ((FILE), "%U%s", (NAME)); \ } while (0) +#undef ASM_OUTPUT_OPCODE +#define ASM_OUTPUT_OPCODE(STREAM, PTR) \ + (PTR) = riscv_asm_output_opcode(STREAM, PTR) + #define JUMP_TABLES_IN_TEXT_SECTION 0 #define CASE_VECTOR_MODE SImode #define CASE_VECTOR_PC_RELATIVE (riscv_cmodel != CM_MEDLOW) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c new file mode 100644 index 00000000000..eee727ef6b4 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c @@ -0,0 +1,12 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gc_xtheadvector -mabi=ilp32 -O0" } */ + +#include "riscv_vector.h" + +vint32m1_t +prefix (vint32m1_t vx, vint32m1_t vy, size_t vl) +{ + return __riscv_vadd_vv_i32m1 (vx, vy, vl); +} + +/* { dg-final { scan-assembler {\mth\.v\M} } } */ -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. 2023-12-29 4:19 ` [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector Jun Sha (Joshua) @ 2023-12-31 17:43 ` Jeff Law 2024-01-01 22:57 ` 钟居哲 2024-01-03 6:08 ` Jun Sha (Joshua) 1 sibling, 1 reply; 133+ messages in thread From: Jeff Law @ 2023-12-31 17:43 UTC (permalink / raw) To: Jun Sha (Joshua), gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, christoph.muellner, juzhe.zhong, Jin Ma, Xianmiao Qu On 12/28/23 21:19, Jun Sha (Joshua) wrote: > This patch adds th. prefix to all XTheadVector instructions by > implementing new assembly output functions. We only check the > prefix is 'v', so that no extra attribute is needed. > > gcc/ChangeLog: > > * config/riscv/riscv-protos.h (riscv_asm_output_opcode): > New function to add assembler insn code prefix/suffix. > * config/riscv/riscv.cc (riscv_asm_output_opcode): Likewise. > * config/riscv/riscv.h (ASM_OUTPUT_OPCODE): Likewise. > > Co-authored-by: Jin Ma <jinma@linux.alibaba.com> > Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> > Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> > --- > gcc/config/riscv/riscv-protos.h | 1 + > gcc/config/riscv/riscv.cc | 14 ++++++++++++++ > gcc/config/riscv/riscv.h | 4 ++++ > .../gcc.target/riscv/rvv/xtheadvector/prefix.c | 12 ++++++++++++ > 4 files changed, 31 insertions(+) > create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c > > diff --git a/gcc/config/riscv/riscv-protos.h b/gcc/config/riscv/riscv-protos.h > index 31049ef7523..5ea54b45703 100644 > --- a/gcc/config/riscv/riscv-protos.h > +++ b/gcc/config/riscv/riscv-protos.h > @@ -102,6 +102,7 @@ struct riscv_address_info { > }; > > /* Routines implemented in riscv.cc. */ > +extern const char *riscv_asm_output_opcode (FILE *asm_out_file, const char *p); > extern enum riscv_symbol_type riscv_classify_symbolic_expression (rtx); > extern bool riscv_symbolic_constant_p (rtx, enum riscv_symbol_type *); > extern int riscv_float_const_rtx_index_for_fli (rtx); > diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc > index 0d1cbc5cb5f..ea1d59d9cf2 100644 > --- a/gcc/config/riscv/riscv.cc > +++ b/gcc/config/riscv/riscv.cc > @@ -5636,6 +5636,20 @@ riscv_get_v_regno_alignment (machine_mode mode) > return lmul; > } > > +/* Define ASM_OUTPUT_OPCODE to do anything special before > + emitting an opcode. */ > +const char * > +riscv_asm_output_opcode (FILE *asm_out_file, const char *p) > +{ > + /* We need to add th. prefix to all the xtheadvector > + insturctions here.*/ > + if (TARGET_XTHEADVECTOR && current_output_insn != NULL_RTX && > + p[0] == 'v') > + fputs ("th.", asm_out_file); > + > + return p; Just a formatting nit. The GNU standards break lines before the operator, not after. So if (TARGET_XTHEADVECTOR && current_output_insn != NULL && p[0] == 'v') Note that current_output_insn is "extern rtx_insn *", so use NULL, not NULL_RTX. Neither of these nits require a new version for review. Just fix them. If Juzhe is fine with this, so am I. We can refine it if necessary later. jeff ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. 2023-12-31 17:43 ` Jeff Law @ 2024-01-01 22:57 ` 钟居哲 2024-01-03 2:54 ` Andrew Pinski 2024-01-04 9:15 ` Re：Re: " joshua 0 siblings, 2 replies; 133+ messages in thread From: 钟居哲 @ 2024-01-01 22:57 UTC (permalink / raw) To: Jeff Law, cooper.joshua, gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, Christoph Müllner, jinma, Cooper Qu [-- Attachment #1: Type: text/plain, Size: 3389 bytes --] This is Ok from my side. But before commit this patch, I think we need this patch first: https://gcc.gnu.org/pipermail/gcc-patches/2023-December/641533.html I will be back to work so I will take a look at other patches today. juzhe.zhong@rivai.ai From: Jeff Law Date: 2024-01-01 01:43 To: Jun Sha (Joshua); gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; christoph.muellner; juzhe.zhong; Jin Ma; Xianmiao Qu Subject: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. On 12/28/23 21:19, Jun Sha (Joshua) wrote: > This patch adds th. prefix to all XTheadVector instructions by > implementing new assembly output functions. We only check the > prefix is 'v', so that no extra attribute is needed. > > gcc/ChangeLog: > > * config/riscv/riscv-protos.h (riscv_asm_output_opcode): > New function to add assembler insn code prefix/suffix. > * config/riscv/riscv.cc (riscv_asm_output_opcode): Likewise. > * config/riscv/riscv.h (ASM_OUTPUT_OPCODE): Likewise. > > Co-authored-by: Jin Ma <jinma@linux.alibaba.com> > Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> > Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> > --- > gcc/config/riscv/riscv-protos.h | 1 + > gcc/config/riscv/riscv.cc | 14 ++++++++++++++ > gcc/config/riscv/riscv.h | 4 ++++ > .../gcc.target/riscv/rvv/xtheadvector/prefix.c | 12 ++++++++++++ > 4 files changed, 31 insertions(+) > create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c > > diff --git a/gcc/config/riscv/riscv-protos.h b/gcc/config/riscv/riscv-protos.h > index 31049ef7523..5ea54b45703 100644 > --- a/gcc/config/riscv/riscv-protos.h > +++ b/gcc/config/riscv/riscv-protos.h > @@ -102,6 +102,7 @@ struct riscv_address_info { > }; > > /* Routines implemented in riscv.cc. */ > +extern const char *riscv_asm_output_opcode (FILE *asm_out_file, const char *p); > extern enum riscv_symbol_type riscv_classify_symbolic_expression (rtx); > extern bool riscv_symbolic_constant_p (rtx, enum riscv_symbol_type *); > extern int riscv_float_const_rtx_index_for_fli (rtx); > diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc > index 0d1cbc5cb5f..ea1d59d9cf2 100644 > --- a/gcc/config/riscv/riscv.cc > +++ b/gcc/config/riscv/riscv.cc > @@ -5636,6 +5636,20 @@ riscv_get_v_regno_alignment (machine_mode mode) > return lmul; > } > > +/* Define ASM_OUTPUT_OPCODE to do anything special before > + emitting an opcode. */ > +const char * > +riscv_asm_output_opcode (FILE *asm_out_file, const char *p) > +{ > + /* We need to add th. prefix to all the xtheadvector > + insturctions here.*/ > + if (TARGET_XTHEADVECTOR && current_output_insn != NULL_RTX && > + p[0] == 'v') > + fputs ("th.", asm_out_file); > + > + return p; Just a formatting nit. The GNU standards break lines before the operator, not after. So if (TARGET_XTHEADVECTOR && current_output_insn != NULL && p[0] == 'v') Note that current_output_insn is "extern rtx_insn *", so use NULL, not NULL_RTX. Neither of these nits require a new version for review. Just fix them. If Juzhe is fine with this, so am I. We can refine it if necessary later. jeff ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. 2024-01-01 22:57 ` 钟居哲 @ 2024-01-03 2:54 ` Andrew Pinski 2024-01-03 3:06 ` juzhe.zhong 2024-01-04 9:15 ` Re：Re: " joshua 1 sibling, 1 reply; 133+ messages in thread From: Andrew Pinski @ 2024-01-03 2:54 UTC (permalink / raw) To: 钟居哲 Cc: Jeff Law, cooper.joshua, gcc-patches, jim.wilson.gcc, palmer, andrew, philipp.tomsich, Christoph Müllner, jinma, Cooper Qu On Mon, Jan 1, 2024 at 2:59 PM 钟居哲 <juzhe.zhong@rivai.ai> wrote: > > This is Ok from my side. > But before commit this patch, I think we need this patch first: > https://gcc.gnu.org/pipermail/gcc-patches/2023-December/641533.html > > I will be back to work so I will take a look at other patches today. Note I hate it. It would be better if you use something like `%^' (see `~` for an example of how that works) instead of hacking riscv_asm_output_opcode really. In fact that is how other targets implement this kind of things. Thanks, Andrew PInski > ________________________________ > juzhe.zhong@rivai.ai > > > From: Jeff Law > Date: 2024-01-01 01:43 > To: Jun Sha (Joshua); gcc-patches > CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; christoph.muellner; juzhe.zhong; Jin Ma; Xianmiao Qu > Subject: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. > > > On 12/28/23 21:19, Jun Sha (Joshua) wrote: > > This patch adds th. prefix to all XTheadVector instructions by > > implementing new assembly output functions. We only check the > > prefix is 'v', so that no extra attribute is needed. > > > > gcc/ChangeLog: > > > > * config/riscv/riscv-protos.h (riscv_asm_output_opcode): > > New function to add assembler insn code prefix/suffix. > > * config/riscv/riscv.cc (riscv_asm_output_opcode): Likewise. > > * config/riscv/riscv.h (ASM_OUTPUT_OPCODE): Likewise. > > > > Co-authored-by: Jin Ma <jinma@linux.alibaba.com> > > Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> > > Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> > > --- > > gcc/config/riscv/riscv-protos.h | 1 + > > gcc/config/riscv/riscv.cc | 14 ++++++++++++++ > > gcc/config/riscv/riscv.h | 4 ++++ > > .../gcc.target/riscv/rvv/xtheadvector/prefix.c | 12 ++++++++++++ > > 4 files changed, 31 insertions(+) > > create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c > > > > diff --git a/gcc/config/riscv/riscv-protos.h b/gcc/config/riscv/riscv-protos.h > > index 31049ef7523..5ea54b45703 100644 > > --- a/gcc/config/riscv/riscv-protos.h > > +++ b/gcc/config/riscv/riscv-protos.h > > @@ -102,6 +102,7 @@ struct riscv_address_info { > > }; > > > > /* Routines implemented in riscv.cc. */ > > +extern const char *riscv_asm_output_opcode (FILE *asm_out_file, const char *p); > > extern enum riscv_symbol_type riscv_classify_symbolic_expression (rtx); > > extern bool riscv_symbolic_constant_p (rtx, enum riscv_symbol_type *); > > extern int riscv_float_const_rtx_index_for_fli (rtx); > > diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc > > index 0d1cbc5cb5f..ea1d59d9cf2 100644 > > --- a/gcc/config/riscv/riscv.cc > > +++ b/gcc/config/riscv/riscv.cc > > @@ -5636,6 +5636,20 @@ riscv_get_v_regno_alignment (machine_mode mode) > > return lmul; > > } > > > > +/* Define ASM_OUTPUT_OPCODE to do anything special before > > + emitting an opcode. */ > > +const char * > > +riscv_asm_output_opcode (FILE *asm_out_file, const char *p) > > +{ > > + /* We need to add th. prefix to all the xtheadvector > > + insturctions here.*/ > > + if (TARGET_XTHEADVECTOR && current_output_insn != NULL_RTX && > > + p[0] == 'v') > > + fputs ("th.", asm_out_file); > > + > > + return p; > Just a formatting nit. The GNU standards break lines before the > operator, not after. So > if (TARGET_XTHEADVECTOR > && current_output_insn != NULL > && p[0] == 'v') > > Note that current_output_insn is "extern rtx_insn *", so use NULL, not > NULL_RTX. > > Neither of these nits require a new version for review. Just fix them. > > If Juzhe is fine with this, so am I. We can refine it if necessary later. > > jeff > ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. 2024-01-03 2:54 ` Andrew Pinski @ 2024-01-03 3:06 ` juzhe.zhong 2024-01-03 3:19 ` Andrew Pinski 0 siblings, 1 reply; 133+ messages in thread From: juzhe.zhong @ 2024-01-03 3:06 UTC (permalink / raw) To: pinskia Cc: jeffreyalaw, cooper.joshua, gcc-patches, Jim Wilson, palmer, andrew, philipp.tomsich, christoph.muellner, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 4391 bytes --] We have no choice. You should know theadvector is totally unrelated with RVV1.0 standard ISA. Adding `%^' which missing totally unrelated ISA makes no sens to me. juzhe.zhong@rivai.ai From: Andrew Pinski Date: 2024-01-03 10:54 To: 钟居哲 CC: Jeff Law; cooper.joshua; gcc-patches; jim.wilson.gcc; palmer; andrew; philipp.tomsich; Christoph Müllner; jinma; Cooper Qu Subject: Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. On Mon, Jan 1, 2024 at 2:59 PM 钟居哲 <juzhe.zhong@rivai.ai> wrote: > > This is Ok from my side. > But before commit this patch, I think we need this patch first: > https://gcc.gnu.org/pipermail/gcc-patches/2023-December/641533.html > > I will be back to work so I will take a look at other patches today. Note I hate it. It would be better if you use something like `%^' (see `~` for an example of how that works) instead of hacking riscv_asm_output_opcode really. In fact that is how other targets implement this kind of things. Thanks, Andrew PInski > ________________________________ > juzhe.zhong@rivai.ai > > > From: Jeff Law > Date: 2024-01-01 01:43 > To: Jun Sha (Joshua); gcc-patches > CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; christoph.muellner; juzhe.zhong; Jin Ma; Xianmiao Qu > Subject: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. > > > On 12/28/23 21:19, Jun Sha (Joshua) wrote: > > This patch adds th. prefix to all XTheadVector instructions by > > implementing new assembly output functions. We only check the > > prefix is 'v', so that no extra attribute is needed. > > > > gcc/ChangeLog: > > > > * config/riscv/riscv-protos.h (riscv_asm_output_opcode): > > New function to add assembler insn code prefix/suffix. > > * config/riscv/riscv.cc (riscv_asm_output_opcode): Likewise. > > * config/riscv/riscv.h (ASM_OUTPUT_OPCODE): Likewise. > > > > Co-authored-by: Jin Ma <jinma@linux.alibaba.com> > > Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> > > Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> > > --- > > gcc/config/riscv/riscv-protos.h | 1 + > > gcc/config/riscv/riscv.cc | 14 ++++++++++++++ > > gcc/config/riscv/riscv.h | 4 ++++ > > .../gcc.target/riscv/rvv/xtheadvector/prefix.c | 12 ++++++++++++ > > 4 files changed, 31 insertions(+) > > create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c > > > > diff --git a/gcc/config/riscv/riscv-protos.h b/gcc/config/riscv/riscv-protos.h > > index 31049ef7523..5ea54b45703 100644 > > --- a/gcc/config/riscv/riscv-protos.h > > +++ b/gcc/config/riscv/riscv-protos.h > > @@ -102,6 +102,7 @@ struct riscv_address_info { > > }; > > > > /* Routines implemented in riscv.cc. */ > > +extern const char *riscv_asm_output_opcode (FILE *asm_out_file, const char *p); > > extern enum riscv_symbol_type riscv_classify_symbolic_expression (rtx); > > extern bool riscv_symbolic_constant_p (rtx, enum riscv_symbol_type *); > > extern int riscv_float_const_rtx_index_for_fli (rtx); > > diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc > > index 0d1cbc5cb5f..ea1d59d9cf2 100644 > > --- a/gcc/config/riscv/riscv.cc > > +++ b/gcc/config/riscv/riscv.cc > > @@ -5636,6 +5636,20 @@ riscv_get_v_regno_alignment (machine_mode mode) > > return lmul; > > } > > > > +/* Define ASM_OUTPUT_OPCODE to do anything special before > > + emitting an opcode. */ > > +const char * > > +riscv_asm_output_opcode (FILE *asm_out_file, const char *p) > > +{ > > + /* We need to add th. prefix to all the xtheadvector > > + insturctions here.*/ > > + if (TARGET_XTHEADVECTOR && current_output_insn != NULL_RTX && > > + p[0] == 'v') > > + fputs ("th.", asm_out_file); > > + > > + return p; > Just a formatting nit. The GNU standards break lines before the > operator, not after. So > if (TARGET_XTHEADVECTOR > && current_output_insn != NULL > && p[0] == 'v') > > Note that current_output_insn is "extern rtx_insn *", so use NULL, not > NULL_RTX. > > Neither of these nits require a new version for review. Just fix them. > > If Juzhe is fine with this, so am I. We can refine it if necessary later. > > jeff > ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. 2024-01-03 3:06 ` juzhe.zhong @ 2024-01-03 3:19 ` Andrew Pinski 2024-01-03 3:25 ` juzhe.zhong 0 siblings, 1 reply; 133+ messages in thread From: Andrew Pinski @ 2024-01-03 3:19 UTC (permalink / raw) To: juzhe.zhong Cc: jeffreyalaw, cooper.joshua, gcc-patches, Jim Wilson, palmer, andrew, philipp.tomsich, christoph.muellner, jinma, cooper.qu On Tue, Jan 2, 2024 at 7:07 PM juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> wrote: > > We have no choice. You should know theadvector is totally unrelated with RVV1.0 standard ISA. > > Adding `%^' which missing totally unrelated ISA makes no sens to me. No, it implements it in a different way. Basically all of the patterns which are supported get changed to be instead of "v*" becomes instead "%^v" and then you change riscv_print_operand_punct_valid_p to allow '^' and then you add '^' support to riscv_print_operand (like '~' is handled there). And the next patch adds a few more '%' to support printing different different strings based on XTheadVector or not. This is how almost all other targets handle this kind of things instead of hacking ASM_OUTPUT_OPCODE . Thanks, Andrew Pinski > > ________________________________ > juzhe.zhong@rivai.ai > > > From: Andrew Pinski > Date: 2024-01-03 10:54 > To: 钟居哲 > CC: Jeff Law; cooper.joshua; gcc-patches; jim.wilson.gcc; palmer; andrew; philipp.tomsich; Christoph Müllner; jinma; Cooper Qu > Subject: Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. > On Mon, Jan 1, 2024 at 2:59 PM 钟居哲 <juzhe.zhong@rivai.ai> wrote: > > > > This is Ok from my side. > > But before commit this patch, I think we need this patch first: > > https://gcc.gnu.org/pipermail/gcc-patches/2023-December/641533.html > > > > I will be back to work so I will take a look at other patches today. > > > Note I hate it. It would be better if you use something like `%^' (see > `~` for an example of how that works) instead of hacking > riscv_asm_output_opcode really. In fact that is how other targets > implement this kind of things. > > Thanks, > Andrew PInski > > > ________________________________ > > juzhe.zhong@rivai.ai > > > > > > From: Jeff Law > > Date: 2024-01-01 01:43 > > To: Jun Sha (Joshua); gcc-patches > > CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; christoph.muellner; juzhe.zhong; Jin Ma; Xianmiao Qu > > Subject: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. > > > > > > On 12/28/23 21:19, Jun Sha (Joshua) wrote: > > > This patch adds th. prefix to all XTheadVector instructions by > > > implementing new assembly output functions. We only check the > > > prefix is 'v', so that no extra attribute is needed. > > > > > > gcc/ChangeLog: > > > > > > * config/riscv/riscv-protos.h (riscv_asm_output_opcode): > > > New function to add assembler insn code prefix/suffix. > > > * config/riscv/riscv.cc (riscv_asm_output_opcode): Likewise. > > > * config/riscv/riscv.h (ASM_OUTPUT_OPCODE): Likewise. > > > > > > Co-authored-by: Jin Ma <jinma@linux.alibaba.com> > > > Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> > > > Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> > > > --- > > > gcc/config/riscv/riscv-protos.h | 1 + > > > gcc/config/riscv/riscv.cc | 14 ++++++++++++++ > > > gcc/config/riscv/riscv.h | 4 ++++ > > > .../gcc.target/riscv/rvv/xtheadvector/prefix.c | 12 ++++++++++++ > > > 4 files changed, 31 insertions(+) > > > create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c > > > > > > diff --git a/gcc/config/riscv/riscv-protos.h b/gcc/config/riscv/riscv-protos.h > > > index 31049ef7523..5ea54b45703 100644 > > > --- a/gcc/config/riscv/riscv-protos.h > > > +++ b/gcc/config/riscv/riscv-protos.h > > > @@ -102,6 +102,7 @@ struct riscv_address_info { > > > }; > > > > > > /* Routines implemented in riscv.cc. */ > > > +extern const char *riscv_asm_output_opcode (FILE *asm_out_file, const char *p); > > > extern enum riscv_symbol_type riscv_classify_symbolic_expression (rtx); > > > extern bool riscv_symbolic_constant_p (rtx, enum riscv_symbol_type *); > > > extern int riscv_float_const_rtx_index_for_fli (rtx); > > > diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc > > > index 0d1cbc5cb5f..ea1d59d9cf2 100644 > > > --- a/gcc/config/riscv/riscv.cc > > > +++ b/gcc/config/riscv/riscv.cc > > > @@ -5636,6 +5636,20 @@ riscv_get_v_regno_alignment (machine_mode mode) > > > return lmul; > > > } > > > > > > +/* Define ASM_OUTPUT_OPCODE to do anything special before > > > + emitting an opcode. */ > > > +const char * > > > +riscv_asm_output_opcode (FILE *asm_out_file, const char *p) > > > +{ > > > + /* We need to add th. prefix to all the xtheadvector > > > + insturctions here.*/ > > > + if (TARGET_XTHEADVECTOR && current_output_insn != NULL_RTX && > > > + p[0] == 'v') > > > + fputs ("th.", asm_out_file); > > > + > > > + return p; > > Just a formatting nit. The GNU standards break lines before the > > operator, not after. So > > if (TARGET_XTHEADVECTOR > > && current_output_insn != NULL > > && p[0] == 'v') > > > > Note that current_output_insn is "extern rtx_insn *", so use NULL, not > > NULL_RTX. > > > > Neither of these nits require a new version for review. Just fix them. > > > > If Juzhe is fine with this, so am I. We can refine it if necessary later. > > > > jeff > > > ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. 2024-01-03 3:19 ` Andrew Pinski @ 2024-01-03 3:25 ` juzhe.zhong 2024-01-03 3:32 ` Andrew Pinski 2024-01-03 3:32 ` Kito Cheng 0 siblings, 2 replies; 133+ messages in thread From: juzhe.zhong @ 2024-01-03 3:25 UTC (permalink / raw) To: pinskia Cc: jeffreyalaw, cooper.joshua, gcc-patches, Jim Wilson, palmer, andrew, philipp.tomsich, christoph.muellner, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 5811 bytes --] No. It will need to change all patterns in vector.md. It's a nightmare. You should note I will refine vector.md in GCC-15, mixing theadvector things make me impossible to maintain RVV1.0. juzhe.zhong@rivai.ai From: Andrew Pinski Date: 2024-01-03 11:19 To: juzhe.zhong@rivai.ai CC: jeffreyalaw; cooper.joshua; gcc-patches; Jim Wilson; palmer; andrew; philipp.tomsich; christoph.muellner; jinma; cooper.qu Subject: Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. On Tue, Jan 2, 2024 at 7:07 PM juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> wrote: > > We have no choice. You should know theadvector is totally unrelated with RVV1.0 standard ISA. > > Adding `%^' which missing totally unrelated ISA makes no sens to me. No, it implements it in a different way. Basically all of the patterns which are supported get changed to be instead of "v*" becomes instead "%^v" and then you change riscv_print_operand_punct_valid_p to allow '^' and then you add '^' support to riscv_print_operand (like '~' is handled there). And the next patch adds a few more '%' to support printing different different strings based on XTheadVector or not. This is how almost all other targets handle this kind of things instead of hacking ASM_OUTPUT_OPCODE . Thanks, Andrew Pinski > > ________________________________ > juzhe.zhong@rivai.ai > > > From: Andrew Pinski > Date: 2024-01-03 10:54 > To: 钟居哲 > CC: Jeff Law; cooper.joshua; gcc-patches; jim.wilson.gcc; palmer; andrew; philipp.tomsich; Christoph Müllner; jinma; Cooper Qu > Subject: Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. > On Mon, Jan 1, 2024 at 2:59 PM 钟居哲 <juzhe.zhong@rivai.ai> wrote: > > > > This is Ok from my side. > > But before commit this patch, I think we need this patch first: > > https://gcc.gnu.org/pipermail/gcc-patches/2023-December/641533.html > > > > I will be back to work so I will take a look at other patches today. > > > Note I hate it. It would be better if you use something like `%^' (see > `~` for an example of how that works) instead of hacking > riscv_asm_output_opcode really. In fact that is how other targets > implement this kind of things. > > Thanks, > Andrew PInski > > > ________________________________ > > juzhe.zhong@rivai.ai > > > > > > From: Jeff Law > > Date: 2024-01-01 01:43 > > To: Jun Sha (Joshua); gcc-patches > > CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; christoph.muellner; juzhe.zhong; Jin Ma; Xianmiao Qu > > Subject: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. > > > > > > On 12/28/23 21:19, Jun Sha (Joshua) wrote: > > > This patch adds th. prefix to all XTheadVector instructions by > > > implementing new assembly output functions. We only check the > > > prefix is 'v', so that no extra attribute is needed. > > > > > > gcc/ChangeLog: > > > > > > * config/riscv/riscv-protos.h (riscv_asm_output_opcode): > > > New function to add assembler insn code prefix/suffix. > > > * config/riscv/riscv.cc (riscv_asm_output_opcode): Likewise. > > > * config/riscv/riscv.h (ASM_OUTPUT_OPCODE): Likewise. > > > > > > Co-authored-by: Jin Ma <jinma@linux.alibaba.com> > > > Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> > > > Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> > > > --- > > > gcc/config/riscv/riscv-protos.h | 1 + > > > gcc/config/riscv/riscv.cc | 14 ++++++++++++++ > > > gcc/config/riscv/riscv.h | 4 ++++ > > > .../gcc.target/riscv/rvv/xtheadvector/prefix.c | 12 ++++++++++++ > > > 4 files changed, 31 insertions(+) > > > create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c > > > > > > diff --git a/gcc/config/riscv/riscv-protos.h b/gcc/config/riscv/riscv-protos.h > > > index 31049ef7523..5ea54b45703 100644 > > > --- a/gcc/config/riscv/riscv-protos.h > > > +++ b/gcc/config/riscv/riscv-protos.h > > > @@ -102,6 +102,7 @@ struct riscv_address_info { > > > }; > > > > > > /* Routines implemented in riscv.cc. */ > > > +extern const char *riscv_asm_output_opcode (FILE *asm_out_file, const char *p); > > > extern enum riscv_symbol_type riscv_classify_symbolic_expression (rtx); > > > extern bool riscv_symbolic_constant_p (rtx, enum riscv_symbol_type *); > > > extern int riscv_float_const_rtx_index_for_fli (rtx); > > > diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc > > > index 0d1cbc5cb5f..ea1d59d9cf2 100644 > > > --- a/gcc/config/riscv/riscv.cc > > > +++ b/gcc/config/riscv/riscv.cc > > > @@ -5636,6 +5636,20 @@ riscv_get_v_regno_alignment (machine_mode mode) > > > return lmul; > > > } > > > > > > +/* Define ASM_OUTPUT_OPCODE to do anything special before > > > + emitting an opcode. */ > > > +const char * > > > +riscv_asm_output_opcode (FILE *asm_out_file, const char *p) > > > +{ > > > + /* We need to add th. prefix to all the xtheadvector > > > + insturctions here.*/ > > > + if (TARGET_XTHEADVECTOR && current_output_insn != NULL_RTX && > > > + p[0] == 'v') > > > + fputs ("th.", asm_out_file); > > > + > > > + return p; > > Just a formatting nit. The GNU standards break lines before the > > operator, not after. So > > if (TARGET_XTHEADVECTOR > > && current_output_insn != NULL > > && p[0] == 'v') > > > > Note that current_output_insn is "extern rtx_insn *", so use NULL, not > > NULL_RTX. > > > > Neither of these nits require a new version for review. Just fix them. > > > > If Juzhe is fine with this, so am I. We can refine it if necessary later. > > > > jeff > > > ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. 2024-01-03 3:25 ` juzhe.zhong @ 2024-01-03 3:32 ` Andrew Pinski 2024-01-03 3:32 ` Kito Cheng 1 sibling, 0 replies; 133+ messages in thread From: Andrew Pinski @ 2024-01-03 3:32 UTC (permalink / raw) To: juzhe.zhong Cc: jeffreyalaw, cooper.joshua, gcc-patches, Jim Wilson, palmer, andrew, philipp.tomsich, christoph.muellner, jinma, cooper.qu On Tue, Jan 2, 2024 at 7:26 PM juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> wrote: > > No. It will need to change all patterns in vector.md. > It's a nightmare. > > You should note I will refine vector.md in GCC-15, mixing theadvector things make me impossible to maintain > RVV1.0. Then we should not support theadvector if things are getting this messy. Both ways are hacks really. Either way we need a better way of implementing this. Hacking theadvector support using rewriting is wrong and not maintainable either. I suspect we should wait on supporting theadvector until GCC 15 anyways. Thanks, Andrew Pinski > > ________________________________ > juzhe.zhong@rivai.ai > > > From: Andrew Pinski > Date: 2024-01-03 11:19 > To: juzhe.zhong@rivai.ai > CC: jeffreyalaw; cooper.joshua; gcc-patches; Jim Wilson; palmer; andrew; philipp.tomsich; christoph.muellner; jinma; cooper.qu > Subject: Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. > On Tue, Jan 2, 2024 at 7:07 PM juzhe.zhong@rivai.ai > <juzhe.zhong@rivai.ai> wrote: > > > > We have no choice. You should know theadvector is totally unrelated with RVV1.0 standard ISA. > > > > Adding `%^' which missing totally unrelated ISA makes no sens to me. > > No, it implements it in a different way. > Basically all of the patterns which are supported get changed to be > instead of "v*" becomes instead "%^v" and then you change > riscv_print_operand_punct_valid_p to allow '^' and then you add '^' > support to riscv_print_operand (like '~' is handled there). > > And the next patch adds a few more '%' to support printing different > different strings based on XTheadVector or not. > > This is how almost all other targets handle this kind of things > instead of hacking ASM_OUTPUT_OPCODE . > > Thanks, > Andrew Pinski > > > > > > ________________________________ > > juzhe.zhong@rivai.ai > > > > > > From: Andrew Pinski > > Date: 2024-01-03 10:54 > > To: 钟居哲 > > CC: Jeff Law; cooper.joshua; gcc-patches; jim.wilson.gcc; palmer; andrew; philipp.tomsich; Christoph Müllner; jinma; Cooper Qu > > Subject: Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. > > On Mon, Jan 1, 2024 at 2:59 PM 钟居哲 <juzhe.zhong@rivai.ai> wrote: > > > > > > This is Ok from my side. > > > But before commit this patch, I think we need this patch first: > > > https://gcc.gnu.org/pipermail/gcc-patches/2023-December/641533.html > > > > > > I will be back to work so I will take a look at other patches today. > > > > > > Note I hate it. It would be better if you use something like `%^' (see > > `~` for an example of how that works) instead of hacking > > riscv_asm_output_opcode really. In fact that is how other targets > > implement this kind of things. > > > > Thanks, > > Andrew PInski > > > > > ________________________________ > > > juzhe.zhong@rivai.ai > > > > > > > > > From: Jeff Law > > > Date: 2024-01-01 01:43 > > > To: Jun Sha (Joshua); gcc-patches > > > CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; christoph.muellner; juzhe.zhong; Jin Ma; Xianmiao Qu > > > Subject: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. > > > > > > > > > On 12/28/23 21:19, Jun Sha (Joshua) wrote: > > > > This patch adds th. prefix to all XTheadVector instructions by > > > > implementing new assembly output functions. We only check the > > > > prefix is 'v', so that no extra attribute is needed. > > > > > > > > gcc/ChangeLog: > > > > > > > > * config/riscv/riscv-protos.h (riscv_asm_output_opcode): > > > > New function to add assembler insn code prefix/suffix. > > > > * config/riscv/riscv.cc (riscv_asm_output_opcode): Likewise. > > > > * config/riscv/riscv.h (ASM_OUTPUT_OPCODE): Likewise. > > > > > > > > Co-authored-by: Jin Ma <jinma@linux.alibaba.com> > > > > Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> > > > > Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> > > > > --- > > > > gcc/config/riscv/riscv-protos.h | 1 + > > > > gcc/config/riscv/riscv.cc | 14 ++++++++++++++ > > > > gcc/config/riscv/riscv.h | 4 ++++ > > > > .../gcc.target/riscv/rvv/xtheadvector/prefix.c | 12 ++++++++++++ > > > > 4 files changed, 31 insertions(+) > > > > create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c > > > > > > > > diff --git a/gcc/config/riscv/riscv-protos.h b/gcc/config/riscv/riscv-protos.h > > > > index 31049ef7523..5ea54b45703 100644 > > > > --- a/gcc/config/riscv/riscv-protos.h > > > > +++ b/gcc/config/riscv/riscv-protos.h > > > > @@ -102,6 +102,7 @@ struct riscv_address_info { > > > > }; > > > > > > > > /* Routines implemented in riscv.cc. */ > > > > +extern const char *riscv_asm_output_opcode (FILE *asm_out_file, const char *p); > > > > extern enum riscv_symbol_type riscv_classify_symbolic_expression (rtx); > > > > extern bool riscv_symbolic_constant_p (rtx, enum riscv_symbol_type *); > > > > extern int riscv_float_const_rtx_index_for_fli (rtx); > > > > diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc > > > > index 0d1cbc5cb5f..ea1d59d9cf2 100644 > > > > --- a/gcc/config/riscv/riscv.cc > > > > +++ b/gcc/config/riscv/riscv.cc > > > > @@ -5636,6 +5636,20 @@ riscv_get_v_regno_alignment (machine_mode mode) > > > > return lmul; > > > > } > > > > > > > > +/* Define ASM_OUTPUT_OPCODE to do anything special before > > > > + emitting an opcode. */ > > > > +const char * > > > > +riscv_asm_output_opcode (FILE *asm_out_file, const char *p) > > > > +{ > > > > + /* We need to add th. prefix to all the xtheadvector > > > > + insturctions here.*/ > > > > + if (TARGET_XTHEADVECTOR && current_output_insn != NULL_RTX && > > > > + p[0] == 'v') > > > > + fputs ("th.", asm_out_file); > > > > + > > > > + return p; > > > Just a formatting nit. The GNU standards break lines before the > > > operator, not after. So > > > if (TARGET_XTHEADVECTOR > > > && current_output_insn != NULL > > > && p[0] == 'v') > > > > > > Note that current_output_insn is "extern rtx_insn *", so use NULL, not > > > NULL_RTX. > > > > > > Neither of these nits require a new version for review. Just fix them. > > > > > > If Juzhe is fine with this, so am I. We can refine it if necessary later. > > > > > > jeff > > > > > > ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. 2024-01-03 3:25 ` juzhe.zhong 2024-01-03 3:32 ` Andrew Pinski @ 2024-01-03 3:32 ` Kito Cheng 1 sibling, 0 replies; 133+ messages in thread From: Kito Cheng @ 2024-01-03 3:32 UTC (permalink / raw) To: juzhe.zhong Cc: pinskia, jeffreyalaw, cooper.joshua, gcc-patches, Jim Wilson, palmer, andrew, philipp.tomsich, christoph.muellner, jinma, cooper.qu Hi Andrew: That's kinda compromise and trade off on the t-head vector stuffs, we would like to accept that, but without disturbing the vector 1.0 implementation too much, t-head vector is transitional product and it will freeze/stop there forever without extending new stuffs like vector bfloat and vector crypto stuffs. So we think using ASM_OUTPUT_OPCODE is better in this case rather than adding %^ to every vector pattern for the t-head vector. On Wed, Jan 3, 2024 at 11:26 AM juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> wrote: > > No. It will need to change all patterns in vector.md. > It's a nightmare. > > You should note I will refine vector.md in GCC-15, mixing theadvector things make me impossible to maintain > RVV1.0. > > ________________________________ > juzhe.zhong@rivai.ai > > > From: Andrew Pinski > Date: 2024-01-03 11:19 > To: juzhe.zhong@rivai.ai > CC: jeffreyalaw; cooper.joshua; gcc-patches; Jim Wilson; palmer; andrew; philipp.tomsich; christoph.muellner; jinma; cooper.qu > Subject: Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. > On Tue, Jan 2, 2024 at 7:07 PM juzhe.zhong@rivai.ai > <juzhe.zhong@rivai.ai> wrote: > > > > We have no choice. You should know theadvector is totally unrelated with RVV1.0 standard ISA. > > > > Adding `%^' which missing totally unrelated ISA makes no sens to me. > > No, it implements it in a different way. > Basically all of the patterns which are supported get changed to be > instead of "v*" becomes instead "%^v" and then you change > riscv_print_operand_punct_valid_p to allow '^' and then you add '^' > support to riscv_print_operand (like '~' is handled there). > > And the next patch adds a few more '%' to support printing different > different strings based on XTheadVector or not. > > This is how almost all other targets handle this kind of things > instead of hacking ASM_OUTPUT_OPCODE . > > Thanks, > Andrew Pinski > > > > > > ________________________________ > > juzhe.zhong@rivai.ai > > > > > > From: Andrew Pinski > > Date: 2024-01-03 10:54 > > To: 钟居哲 > > CC: Jeff Law; cooper.joshua; gcc-patches; jim.wilson.gcc; palmer; andrew; philipp.tomsich; Christoph Müllner; jinma; Cooper Qu > > Subject: Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. > > On Mon, Jan 1, 2024 at 2:59 PM 钟居哲 <juzhe.zhong@rivai.ai> wrote: > > > > > > This is Ok from my side. > > > But before commit this patch, I think we need this patch first: > > > https://gcc.gnu.org/pipermail/gcc-patches/2023-December/641533.html > > > > > > I will be back to work so I will take a look at other patches today. > > > > > > Note I hate it. It would be better if you use something like `%^' (see > > `~` for an example of how that works) instead of hacking > > riscv_asm_output_opcode really. In fact that is how other targets > > implement this kind of things. > > > > Thanks, > > Andrew PInski > > > > > ________________________________ > > > juzhe.zhong@rivai.ai > > > > > > > > > From: Jeff Law > > > Date: 2024-01-01 01:43 > > > To: Jun Sha (Joshua); gcc-patches > > > CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; christoph.muellner; juzhe.zhong; Jin Ma; Xianmiao Qu > > > Subject: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. > > > > > > > > > On 12/28/23 21:19, Jun Sha (Joshua) wrote: > > > > This patch adds th. prefix to all XTheadVector instructions by > > > > implementing new assembly output functions. We only check the > > > > prefix is 'v', so that no extra attribute is needed. > > > > > > > > gcc/ChangeLog: > > > > > > > > * config/riscv/riscv-protos.h (riscv_asm_output_opcode): > > > > New function to add assembler insn code prefix/suffix. > > > > * config/riscv/riscv.cc (riscv_asm_output_opcode): Likewise. > > > > * config/riscv/riscv.h (ASM_OUTPUT_OPCODE): Likewise. > > > > > > > > Co-authored-by: Jin Ma <jinma@linux.alibaba.com> > > > > Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> > > > > Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> > > > > --- > > > > gcc/config/riscv/riscv-protos.h | 1 + > > > > gcc/config/riscv/riscv.cc | 14 ++++++++++++++ > > > > gcc/config/riscv/riscv.h | 4 ++++ > > > > .../gcc.target/riscv/rvv/xtheadvector/prefix.c | 12 ++++++++++++ > > > > 4 files changed, 31 insertions(+) > > > > create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c > > > > > > > > diff --git a/gcc/config/riscv/riscv-protos.h b/gcc/config/riscv/riscv-protos.h > > > > index 31049ef7523..5ea54b45703 100644 > > > > --- a/gcc/config/riscv/riscv-protos.h > > > > +++ b/gcc/config/riscv/riscv-protos.h > > > > @@ -102,6 +102,7 @@ struct riscv_address_info { > > > > }; > > > > > > > > /* Routines implemented in riscv.cc. */ > > > > +extern const char *riscv_asm_output_opcode (FILE *asm_out_file, const char *p); > > > > extern enum riscv_symbol_type riscv_classify_symbolic_expression (rtx); > > > > extern bool riscv_symbolic_constant_p (rtx, enum riscv_symbol_type *); > > > > extern int riscv_float_const_rtx_index_for_fli (rtx); > > > > diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc > > > > index 0d1cbc5cb5f..ea1d59d9cf2 100644 > > > > --- a/gcc/config/riscv/riscv.cc > > > > +++ b/gcc/config/riscv/riscv.cc > > > > @@ -5636,6 +5636,20 @@ riscv_get_v_regno_alignment (machine_mode mode) > > > > return lmul; > > > > } > > > > > > > > +/* Define ASM_OUTPUT_OPCODE to do anything special before > > > > + emitting an opcode. */ > > > > +const char * > > > > +riscv_asm_output_opcode (FILE *asm_out_file, const char *p) > > > > +{ > > > > + /* We need to add th. prefix to all the xtheadvector > > > > + insturctions here.*/ > > > > + if (TARGET_XTHEADVECTOR && current_output_insn != NULL_RTX && > > > > + p[0] == 'v') > > > > + fputs ("th.", asm_out_file); > > > > + > > > > + return p; > > > Just a formatting nit. The GNU standards break lines before the > > > operator, not after. So > > > if (TARGET_XTHEADVECTOR > > > && current_output_insn != NULL > > > && p[0] == 'v') > > > > > > Note that current_output_insn is "extern rtx_insn *", so use NULL, not > > > NULL_RTX. > > > > > > Neither of these nits require a new version for review. Just fix them. > > > > > > If Juzhe is fine with this, so am I. We can refine it if necessary later. > > > > > > jeff > > > > > > ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re：Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. 2024-01-01 22:57 ` 钟居哲 2024-01-03 2:54 ` Andrew Pinski @ 2024-01-04 9:15 ` joshua 2024-01-04 9:18 ` juzhe.zhong 1 sibling, 1 reply; 133+ messages in thread From: joshua @ 2024-01-04 9:15 UTC (permalink / raw) To: 钟居哲, Jeff Law, gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, Christoph Müllner, jinma, Cooper Qu Hi Juzhe, So is the following patch that this patch relies on OK to commit? https://gcc.gnu.org/pipermail/gcc-patches/2023-December/641533.html Joshua ------------------------------------------------------------------ 发件人：钟居哲 <juzhe.zhong@rivai.ai> 发送时间：2024年1月2日(星期二) 06:57 收件人：Jeff Law<jeffreyalaw@gmail.com>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送："jim.wilson.gcc"<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; "Christoph Müllner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; Cooper Qu<cooper.qu@linux.alibaba.com> 主 题：Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. This is Ok from my side. But before commit this patch, I think we need this patch first: https://gcc.gnu.org/pipermail/gcc-patches/2023-December/641533.html I will be back to work so I will take a look at other patches today. juzhe.zhong@rivai.ai From: Jeff Law Date: 2024-01-01 01:43 To: Jun Sha (Joshua); gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; christoph.muellner; juzhe.zhong; Jin Ma; Xianmiao Qu Subject: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. On 12/28/23 21:19, Jun Sha (Joshua) wrote: > This patch adds th. prefix to all XTheadVector instructions by > implementing new assembly output functions. We only check the > prefix is 'v', so that no extra attribute is needed. > > gcc/ChangeLog: > > * config/riscv/riscv-protos.h (riscv_asm_output_opcode): > New function to add assembler insn code prefix/suffix. > * config/riscv/riscv.cc (riscv_asm_output_opcode): Likewise. > * config/riscv/riscv.h (ASM_OUTPUT_OPCODE): Likewise. > > Co-authored-by: Jin Ma <jinma@linux.alibaba.com> > Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> > Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> > --- > gcc/config/riscv/riscv-protos.h | 1 + > gcc/config/riscv/riscv.cc | 14 ++++++++++++++ > gcc/config/riscv/riscv.h | 4 ++++ > .../gcc.target/riscv/rvv/xtheadvector/prefix.c | 12 ++++++++++++ > 4 files changed, 31 insertions(+) > create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c > > diff --git a/gcc/config/riscv/riscv-protos.h b/gcc/config/riscv/riscv-protos.h > index 31049ef7523..5ea54b45703 100644 > --- a/gcc/config/riscv/riscv-protos.h > +++ b/gcc/config/riscv/riscv-protos.h > @@ -102,6 +102,7 @@ struct riscv_address_info { > }; > > /* Routines implemented in riscv.cc. */ > +extern const char *riscv_asm_output_opcode (FILE *asm_out_file, const char *p); > extern enum riscv_symbol_type riscv_classify_symbolic_expression (rtx); > extern bool riscv_symbolic_constant_p (rtx, enum riscv_symbol_type *); > extern int riscv_float_const_rtx_index_for_fli (rtx); > diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc > index 0d1cbc5cb5f..ea1d59d9cf2 100644 > --- a/gcc/config/riscv/riscv.cc > +++ b/gcc/config/riscv/riscv.cc > @@ -5636,6 +5636,20 @@ riscv_get_v_regno_alignment (machine_mode mode) > return lmul; > } > > +/* Define ASM_OUTPUT_OPCODE to do anything special before > + emitting an opcode. */ > +const char * > +riscv_asm_output_opcode (FILE *asm_out_file, const char *p) > +{ > + /* We need to add th. prefix to all the xtheadvector > + insturctions here.*/ > + if (TARGET_XTHEADVECTOR && current_output_insn != NULL_RTX && > + p[0] == 'v') > + fputs ("th.", asm_out_file); > + > + return p; Just a formatting nit. The GNU standards break lines before the operator, not after. So if (TARGET_XTHEADVECTOR && current_output_insn != NULL && p[0] == 'v') Note that current_output_insn is "extern rtx_insn *", so use NULL, not NULL_RTX. Neither of these nits require a new version for review. Just fix them. If Juzhe is fine with this, so am I. We can refine it if necessary later. jeff ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. 2024-01-04 9:15 ` Re：Re: " joshua @ 2024-01-04 9:18 ` juzhe.zhong 2024-01-04 10:04 ` Christoph Müllner 2024-01-08 2:11 ` Re：Re: " joshua 0 siblings, 2 replies; 133+ messages in thread From: juzhe.zhong @ 2024-01-04 9:18 UTC (permalink / raw) To: cooper.joshua, jeffreyalaw, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, christoph.muellner, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 4794 bytes --] \ No newline at end of file Each file needs newline. I am not able to review arch stuff. This needs kito. Besides, Andrew Pinski want us defer theadvector to GCC-15. I have no strong opinion here. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-04 17:15 收件人： 钟居哲; Jeff Law; gcc-patches 抄送： jim.wilson.gcc; palmer; andrew; philipp.tomsich; Christoph Müllner; jinma; Cooper Qu 主题： Re：Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. Hi Juzhe, So is the following patch that this patch relies on OK to commit? https://gcc.gnu.org/pipermail/gcc-patches/2023-December/641533.html Joshua ------------------------------------------------------------------ 发件人：钟居哲 <juzhe.zhong@rivai.ai> 发送时间：2024年1月2日(星期二) 06:57 收件人：Jeff Law<jeffreyalaw@gmail.com>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送："jim.wilson.gcc"<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; "Christoph Müllner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; Cooper Qu<cooper.qu@linux.alibaba.com> 主 题：Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. This is Ok from my side. But before commit this patch, I think we need this patch first: https://gcc.gnu.org/pipermail/gcc-patches/2023-December/641533.html I will be back to work so I will take a look at other patches today. juzhe.zhong@rivai.ai From: Jeff Law Date: 2024-01-01 01:43 To: Jun Sha (Joshua); gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; christoph.muellner; juzhe.zhong; Jin Ma; Xianmiao Qu Subject: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. On 12/28/23 21:19, Jun Sha (Joshua) wrote: > This patch adds th. prefix to all XTheadVector instructions by > implementing new assembly output functions. We only check the > prefix is 'v', so that no extra attribute is needed. > > gcc/ChangeLog: > > * config/riscv/riscv-protos.h (riscv_asm_output_opcode): > New function to add assembler insn code prefix/suffix. > * config/riscv/riscv.cc (riscv_asm_output_opcode): Likewise. > * config/riscv/riscv.h (ASM_OUTPUT_OPCODE): Likewise. > > Co-authored-by: Jin Ma <jinma@linux.alibaba.com> > Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> > Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> > --- > gcc/config/riscv/riscv-protos.h | 1 + > gcc/config/riscv/riscv.cc | 14 ++++++++++++++ > gcc/config/riscv/riscv.h | 4 ++++ > .../gcc.target/riscv/rvv/xtheadvector/prefix.c | 12 ++++++++++++ > 4 files changed, 31 insertions(+) > create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c > > diff --git a/gcc/config/riscv/riscv-protos.h b/gcc/config/riscv/riscv-protos.h > index 31049ef7523..5ea54b45703 100644 > --- a/gcc/config/riscv/riscv-protos.h > +++ b/gcc/config/riscv/riscv-protos.h > @@ -102,6 +102,7 @@ struct riscv_address_info { > }; > > /* Routines implemented in riscv.cc. */ > +extern const char *riscv_asm_output_opcode (FILE *asm_out_file, const char *p); > extern enum riscv_symbol_type riscv_classify_symbolic_expression (rtx); > extern bool riscv_symbolic_constant_p (rtx, enum riscv_symbol_type *); > extern int riscv_float_const_rtx_index_for_fli (rtx); > diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc > index 0d1cbc5cb5f..ea1d59d9cf2 100644 > --- a/gcc/config/riscv/riscv.cc > +++ b/gcc/config/riscv/riscv.cc > @@ -5636,6 +5636,20 @@ riscv_get_v_regno_alignment (machine_mode mode) > return lmul; > } > > +/* Define ASM_OUTPUT_OPCODE to do anything special before > + emitting an opcode. */ > +const char * > +riscv_asm_output_opcode (FILE *asm_out_file, const char *p) > +{ > + /* We need to add th. prefix to all the xtheadvector > + insturctions here.*/ > + if (TARGET_XTHEADVECTOR && current_output_insn != NULL_RTX && > + p[0] == 'v') > + fputs ("th.", asm_out_file); > + > + return p; Just a formatting nit. The GNU standards break lines before the operator, not after. So if (TARGET_XTHEADVECTOR && current_output_insn != NULL && p[0] == 'v') Note that current_output_insn is "extern rtx_insn *", so use NULL, not NULL_RTX. Neither of these nits require a new version for review. Just fix them. If Juzhe is fine with this, so am I. We can refine it if necessary later. jeff ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. 2024-01-04 9:18 ` juzhe.zhong @ 2024-01-04 10:04 ` Christoph Müllner 2024-01-08 2:11 ` Re：Re: " joshua 1 sibling, 0 replies; 133+ messages in thread From: Christoph Müllner @ 2024-01-04 10:04 UTC (permalink / raw) To: juzhe.zhong Cc: cooper.joshua, jeffreyalaw, gcc-patches, Jim Wilson, palmer, andrew, philipp.tomsich, jinma, cooper.qu On Thu, Jan 4, 2024 at 10:18 AM juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> wrote: > > \ No newline at end of file > > Each file needs newline. > > I am not able to review arch stuff. This needs kito. > > Besides, Andrew Pinski want us defer theadvector to GCC-15. Maybe I misread this (sorry if so), but I though that was answered by Kito here: https://gcc.gnu.org/pipermail/gcc-patches/2024-January/641723.html > > I have no strong opinion here. > > ________________________________ > juzhe.zhong@rivai.ai > > > 发件人： joshua > 发送时间： 2024-01-04 17:15 > 收件人： 钟居哲; Jeff Law; gcc-patches > 抄送： jim.wilson.gcc; palmer; andrew; philipp.tomsich; Christoph Müllner; jinma; Cooper Qu > 主题： Re：Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. > Hi Juzhe, > > So is the following patch that this patch relies on OK to commit? > https://gcc.gnu.org/pipermail/gcc-patches/2023-December/641533.html > > Joshua > > > > > ------------------------------------------------------------------ > 发件人：钟居哲 <juzhe.zhong@rivai.ai> > 发送时间：2024年1月2日(星期二) 06:57 > 收件人：Jeff Law<jeffreyalaw@gmail.com>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> > 抄 送："jim.wilson.gcc"<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; "Christoph Müllner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; Cooper Qu<cooper.qu@linux.alibaba.com> > 主 题：Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. > > > This is Ok from my side. > But before commit this patch, I think we need this patch first: > https://gcc.gnu.org/pipermail/gcc-patches/2023-December/641533.html > > > I will be back to work so I will take a look at other patches today. > juzhe.zhong@rivai.ai > > > From: Jeff Law > Date: 2024-01-01 01:43 > To: Jun Sha (Joshua); gcc-patches > CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; christoph.muellner; juzhe.zhong; Jin Ma; Xianmiao Qu > Subject: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. > > > > On 12/28/23 21:19, Jun Sha (Joshua) wrote: > > This patch adds th. prefix to all XTheadVector instructions by > > implementing new assembly output functions. We only check the > > prefix is 'v', so that no extra attribute is needed. > > > > gcc/ChangeLog: > > > > * config/riscv/riscv-protos.h (riscv_asm_output_opcode): > > New function to add assembler insn code prefix/suffix. > > * config/riscv/riscv.cc (riscv_asm_output_opcode): Likewise. > > * config/riscv/riscv.h (ASM_OUTPUT_OPCODE): Likewise. > > > > Co-authored-by: Jin Ma <jinma@linux.alibaba.com> > > Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> > > Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> > > --- > > gcc/config/riscv/riscv-protos.h | 1 + > > gcc/config/riscv/riscv.cc | 14 ++++++++++++++ > > gcc/config/riscv/riscv.h | 4 ++++ > > .../gcc.target/riscv/rvv/xtheadvector/prefix.c | 12 ++++++++++++ > > 4 files changed, 31 insertions(+) > > create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c > > > > diff --git a/gcc/config/riscv/riscv-protos.h b/gcc/config/riscv/riscv-protos.h > > index 31049ef7523..5ea54b45703 100644 > > --- a/gcc/config/riscv/riscv-protos.h > > +++ b/gcc/config/riscv/riscv-protos.h > > @@ -102,6 +102,7 @@ struct riscv_address_info { > > }; > > > > /* Routines implemented in riscv.cc. */ > > +extern const char *riscv_asm_output_opcode (FILE *asm_out_file, const char *p); > > extern enum riscv_symbol_type riscv_classify_symbolic_expression (rtx); > > extern bool riscv_symbolic_constant_p (rtx, enum riscv_symbol_type *); > > extern int riscv_float_const_rtx_index_for_fli (rtx); > > diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc > > index 0d1cbc5cb5f..ea1d59d9cf2 100644 > > --- a/gcc/config/riscv/riscv.cc > > +++ b/gcc/config/riscv/riscv.cc > > @@ -5636,6 +5636,20 @@ riscv_get_v_regno_alignment (machine_mode mode) > > return lmul; > > } > > > > +/* Define ASM_OUTPUT_OPCODE to do anything special before > > + emitting an opcode. */ > > +const char * > > +riscv_asm_output_opcode (FILE *asm_out_file, const char *p) > > +{ > > + /* We need to add th. prefix to all the xtheadvector > > + insturctions here.*/ > > + if (TARGET_XTHEADVECTOR && current_output_insn != NULL_RTX && > > + p[0] == 'v') > > + fputs ("th.", asm_out_file); > > + > > + return p; > Just a formatting nit. The GNU standards break lines before the > operator, not after. So > if (TARGET_XTHEADVECTOR > && current_output_insn != NULL > && p[0] == 'v') > > Note that current_output_insn is "extern rtx_insn *", so use NULL, not > NULL_RTX. > > Neither of these nits require a new version for review. Just fix them. > > If Juzhe is fine with this, so am I. We can refine it if necessary later. > > jeff > > > ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re：Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. 2024-01-04 9:18 ` juzhe.zhong 2024-01-04 10:04 ` Christoph Müllner @ 2024-01-08 2:11 ` joshua 2024-01-08 3:06 ` Kito Cheng 1 sibling, 1 reply; 133+ messages in thread From: joshua @ 2024-01-08 2:11 UTC (permalink / raw) To: juzhe.zhong, jeffreyalaw, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, christoph.muellner, jinma, cooper.qu Hi Juzhe, Stage 3 will close today and there are still some patches that haven't been reviewed left. So is it possible to get xtheadvector merged in GCC-14? We emailed Kito regarding this, but haven't got any reply yet. Joshua ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月4日(星期四) 17:18 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; jeffreyalaw<jeffreyalaw@gmail.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. \ No newline at end of file Each file needs newline. I am not able to review arch stuff. This needs kito. Besides, Andrew Pinski want us defer theadvector to GCC-15. I have no strong opinion here. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-04 17:15 收件人： 钟居哲; Jeff Law; gcc-patches 抄送： jim.wilson.gcc; palmer; andrew; philipp.tomsich; Christoph Müllner; jinma; Cooper Qu 主题： Re：Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. Hi Juzhe, So is the following patch that this patch relies on OK to commit? https://gcc.gnu.org/pipermail/gcc-patches/2023-December/641533.html Joshua ------------------------------------------------------------------ 发件人：钟居哲 <juzhe.zhong@rivai.ai> 发送时间：2024年1月2日(星期二) 06:57 收件人：Jeff Law<jeffreyalaw@gmail.com>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送："jim.wilson.gcc"<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; "Christoph Müllner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; Cooper Qu<cooper.qu@linux.alibaba.com> 主 题：Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. This is Ok from my side. But before commit this patch, I think we need this patch first: https://gcc.gnu.org/pipermail/gcc-patches/2023-December/641533.html I will be back to work so I will take a look at other patches today. juzhe.zhong@rivai.ai From: Jeff Law Date: 2024-01-01 01:43 To: Jun Sha (Joshua); gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; christoph.muellner; juzhe.zhong; Jin Ma; Xianmiao Qu Subject: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. On 12/28/23 21:19, Jun Sha (Joshua) wrote: > This patch adds th. prefix to all XTheadVector instructions by > implementing new assembly output functions. We only check the > prefix is 'v', so that no extra attribute is needed. > > gcc/ChangeLog: > > * config/riscv/riscv-protos.h (riscv_asm_output_opcode): > New function to add assembler insn code prefix/suffix. > * config/riscv/riscv.cc (riscv_asm_output_opcode): Likewise. > * config/riscv/riscv.h (ASM_OUTPUT_OPCODE): Likewise. > > Co-authored-by: Jin Ma <jinma@linux.alibaba.com> > Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> > Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> > --- > gcc/config/riscv/riscv-protos.h | 1 + > gcc/config/riscv/riscv.cc | 14 ++++++++++++++ > gcc/config/riscv/riscv.h | 4 ++++ > .../gcc.target/riscv/rvv/xtheadvector/prefix.c | 12 ++++++++++++ > 4 files changed, 31 insertions(+) > create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c > > diff --git a/gcc/config/riscv/riscv-protos.h b/gcc/config/riscv/riscv-protos.h > index 31049ef7523..5ea54b45703 100644 > --- a/gcc/config/riscv/riscv-protos.h > +++ b/gcc/config/riscv/riscv-protos.h > @@ -102,6 +102,7 @@ struct riscv_address_info { > }; > > /* Routines implemented in riscv.cc. */ > +extern const char *riscv_asm_output_opcode (FILE *asm_out_file, const char *p); > extern enum riscv_symbol_type riscv_classify_symbolic_expression (rtx); > extern bool riscv_symbolic_constant_p (rtx, enum riscv_symbol_type *); > extern int riscv_float_const_rtx_index_for_fli (rtx); > diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc > index 0d1cbc5cb5f..ea1d59d9cf2 100644 > --- a/gcc/config/riscv/riscv.cc > +++ b/gcc/config/riscv/riscv.cc > @@ -5636,6 +5636,20 @@ riscv_get_v_regno_alignment (machine_mode mode) > return lmul; > } > > +/* Define ASM_OUTPUT_OPCODE to do anything special before > + emitting an opcode. */ > +const char * > +riscv_asm_output_opcode (FILE *asm_out_file, const char *p) > +{ > + /* We need to add th. prefix to all the xtheadvector > + insturctions here.*/ > + if (TARGET_XTHEADVECTOR && current_output_insn != NULL_RTX && > + p[0] == 'v') > + fputs ("th.", asm_out_file); > + > + return p; Just a formatting nit. The GNU standards break lines before the operator, not after. So if (TARGET_XTHEADVECTOR && current_output_insn != NULL && p[0] == 'v') Note that current_output_insn is "extern rtx_insn *", so use NULL, not NULL_RTX. Neither of these nits require a new version for review. Just fix them. If Juzhe is fine with this, so am I. We can refine it if necessary later. jeff ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. 2024-01-08 2:11 ` Re：Re: " joshua @ 2024-01-08 3:06 ` Kito Cheng 2024-01-08 3:17 ` Re：Re: " joshua 0 siblings, 1 reply; 133+ messages in thread From: Kito Cheng @ 2024-01-08 3:06 UTC (permalink / raw) To: joshua Cc: juzhe.zhong, jeffreyalaw, gcc-patches, Jim Wilson, palmer, andrew, philipp.tomsich, christoph.muellner, jinma, cooper.qu I am ok with merging this for GCC 14, as we discussed several times in the RISC-V GCC sync up meeting, I think at least we reach consensus among Jeff Law, Palmer Dabbelt and me. But please be careful: don't break anything for standard vector stuff. On Mon, Jan 8, 2024 at 10:11 AM joshua <cooper.joshua@linux.alibaba.com> wrote: > > Hi Juzhe, > > Stage 3 will close today and there are still some patches that > haven't been reviewed left. > So is it possible to get xtheadvector merged in GCC-14? > We emailed Kito regarding this, but haven't got any reply yet. > > Joshua > > > > > > > ------------------------------------------------------------------ > 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> > 发送时间：2024年1月4日(星期四) 17:18 > 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; jeffreyalaw<jeffreyalaw@gmail.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> > 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> > 主 题：Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. > > > \ No newline at end of file > Each file needs newline. > > > I am not able to review arch stuff. This needs kito. > > > Besides, Andrew Pinski want us defer theadvector to GCC-15. > > > I have no strong opinion here. > > > juzhe.zhong@rivai.ai > > > 发件人： joshua > 发送时间： 2024-01-04 17:15 > 收件人： 钟居哲; Jeff Law; gcc-patches > 抄送： jim.wilson.gcc; palmer; andrew; philipp.tomsich; Christoph Müllner; jinma; Cooper Qu > 主题： Re：Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. > > Hi Juzhe, > > So is the following patch that this patch relies on OK to commit? > https://gcc.gnu.org/pipermail/gcc-patches/2023-December/641533.html > > Joshua > > > > > ------------------------------------------------------------------ > 发件人：钟居哲 <juzhe.zhong@rivai.ai> > 发送时间：2024年1月2日(星期二) 06:57 > 收件人：Jeff Law<jeffreyalaw@gmail.com>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> > 抄 送："jim.wilson.gcc"<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; "Christoph Müllner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; Cooper Qu<cooper.qu@linux.alibaba.com> > 主 题：Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. > > > This is Ok from my side. > But before commit this patch, I think we need this patch first: > https://gcc.gnu.org/pipermail/gcc-patches/2023-December/641533.html > > > I will be back to work so I will take a look at other patches today. > juzhe.zhong@rivai.ai > > > From: Jeff Law > Date: 2024-01-01 01:43 > To: Jun Sha (Joshua); gcc-patches > CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; christoph.muellner; juzhe.zhong; Jin Ma; Xianmiao Qu > Subject: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. > > > > On 12/28/23 21:19, Jun Sha (Joshua) wrote: > > This patch adds th. prefix to all XTheadVector instructions by > > implementing new assembly output functions. We only check the > > prefix is 'v', so that no extra attribute is needed. > > > > gcc/ChangeLog: > > > > * config/riscv/riscv-protos.h (riscv_asm_output_opcode): > > New function to add assembler insn code prefix/suffix. > > * config/riscv/riscv.cc (riscv_asm_output_opcode): Likewise. > > * config/riscv/riscv.h (ASM_OUTPUT_OPCODE): Likewise. > > > > Co-authored-by: Jin Ma <jinma@linux.alibaba.com> > > Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> > > Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> > > --- > > gcc/config/riscv/riscv-protos.h | 1 + > > gcc/config/riscv/riscv.cc | 14 ++++++++++++++ > > gcc/config/riscv/riscv.h | 4 ++++ > > .../gcc.target/riscv/rvv/xtheadvector/prefix.c | 12 ++++++++++++ > > 4 files changed, 31 insertions(+) > > create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c > > > > diff --git a/gcc/config/riscv/riscv-protos.h b/gcc/config/riscv/riscv-protos.h > > index 31049ef7523..5ea54b45703 100644 > > --- a/gcc/config/riscv/riscv-protos.h > > +++ b/gcc/config/riscv/riscv-protos.h > > @@ -102,6 +102,7 @@ struct riscv_address_info { > > }; > > > > /* Routines implemented in riscv.cc. */ > > +extern const char *riscv_asm_output_opcode (FILE *asm_out_file, const char *p); > > extern enum riscv_symbol_type riscv_classify_symbolic_expression (rtx); > > extern bool riscv_symbolic_constant_p (rtx, enum riscv_symbol_type *); > > extern int riscv_float_const_rtx_index_for_fli (rtx); > > diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc > > index 0d1cbc5cb5f..ea1d59d9cf2 100644 > > --- a/gcc/config/riscv/riscv.cc > > +++ b/gcc/config/riscv/riscv.cc > > @@ -5636,6 +5636,20 @@ riscv_get_v_regno_alignment (machine_mode mode) > > return lmul; > > } > > > > +/* Define ASM_OUTPUT_OPCODE to do anything special before > > + emitting an opcode. */ > > +const char * > > +riscv_asm_output_opcode (FILE *asm_out_file, const char *p) > > +{ > > + /* We need to add th. prefix to all the xtheadvector > > + insturctions here.*/ > > + if (TARGET_XTHEADVECTOR && current_output_insn != NULL_RTX && > > + p[0] == 'v') > > + fputs ("th.", asm_out_file); > > + > > + return p; > Just a formatting nit. The GNU standards break lines before the > operator, not after. So > if (TARGET_XTHEADVECTOR > && current_output_insn != NULL > && p[0] == 'v') > > Note that current_output_insn is "extern rtx_insn *", so use NULL, not > NULL_RTX. > > Neither of these nits require a new version for review. Just fix them. > > If Juzhe is fine with this, so am I. We can refine it if necessary later. > > jeff > > > > > ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re：Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. 2024-01-08 3:06 ` Kito Cheng @ 2024-01-08 3:17 ` joshua 2024-01-08 3:40 ` Kito Cheng 0 siblings, 1 reply; 133+ messages in thread From: joshua @ 2024-01-08 3:17 UTC (permalink / raw) To: Kito Cheng Cc: juzhe.zhong, jeffreyalaw, gcc-patches, Jim Wilson, palmer, andrew, philipp.tomsich, christoph.muellner, jinma, cooper.qu Hi Kito, Thank you for your support. So even during stage 4, we can merge this for GCC 14? ------------------------------------------------------------------ 发件人：Kito Cheng <kito.cheng@gmail.com> 发送时间：2024年1月8日(星期一) 11:06 收件人：joshua<cooper.joshua@linux.alibaba.com> 抄 送："juzhe.zhong@rivai.ai"<juzhe.zhong@rivai.ai>; jeffreyalaw<jeffreyalaw@gmail.com>; "gcc-patches"<gcc-patches@gcc.gnu.org>; Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. I am ok with merging this for GCC 14, as we discussed several times in the RISC-V GCC sync up meeting, I think at least we reach consensus among Jeff Law, Palmer Dabbelt and me. But please be careful: don't break anything for standard vector stuff. On Mon, Jan 8, 2024 at 10:11 AM joshua <cooper.joshua@linux.alibaba.com> wrote: > > Hi Juzhe, > > Stage 3 will close today and there are still some patches that > haven't been reviewed left. > So is it possible to get xtheadvector merged in GCC-14? > We emailed Kito regarding this, but haven't got any reply yet. > > Joshua > > > > > > > ------------------------------------------------------------------ > 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> > 发送时间：2024年1月4日(星期四) 17:18 > 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; jeffreyalaw<jeffreyalaw@gmail.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> > 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> > 主 题：Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. > > > \ No newline at end of file > Each file needs newline. > > > I am not able to review arch stuff. This needs kito. > > > Besides, Andrew Pinski want us defer theadvector to GCC-15. > > > I have no strong opinion here. > > > juzhe.zhong@rivai.ai > > > 发件人： joshua > 发送时间： 2024-01-04 17:15 > 收件人： 钟居哲; Jeff Law; gcc-patches > 抄送： jim.wilson.gcc; palmer; andrew; philipp.tomsich; Christoph Müllner; jinma; Cooper Qu > 主题： Re：Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. > > Hi Juzhe, > > So is the following patch that this patch relies on OK to commit? > https://gcc.gnu.org/pipermail/gcc-patches/2023-December/641533.html > > Joshua > > > > > ------------------------------------------------------------------ > 发件人：钟居哲 <juzhe.zhong@rivai.ai> > 发送时间：2024年1月2日(星期二) 06:57 > 收件人：Jeff Law<jeffreyalaw@gmail.com>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> > 抄 送："jim.wilson.gcc"<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; "Christoph Müllner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; Cooper Qu<cooper.qu@linux.alibaba.com> > 主 题：Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. > > > This is Ok from my side. > But before commit this patch, I think we need this patch first: > https://gcc.gnu.org/pipermail/gcc-patches/2023-December/641533.html > > > I will be back to work so I will take a look at other patches today. > juzhe.zhong@rivai.ai > > > From: Jeff Law > Date: 2024-01-01 01:43 > To: Jun Sha (Joshua); gcc-patches > CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; christoph.muellner; juzhe.zhong; Jin Ma; Xianmiao Qu > Subject: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. > > > > On 12/28/23 21:19, Jun Sha (Joshua) wrote: > > This patch adds th. prefix to all XTheadVector instructions by > > implementing new assembly output functions. We only check the > > prefix is 'v', so that no extra attribute is needed. > > > > gcc/ChangeLog: > > > > * config/riscv/riscv-protos.h (riscv_asm_output_opcode): > > New function to add assembler insn code prefix/suffix. > > * config/riscv/riscv.cc (riscv_asm_output_opcode): Likewise. > > * config/riscv/riscv.h (ASM_OUTPUT_OPCODE): Likewise. > > > > Co-authored-by: Jin Ma <jinma@linux.alibaba.com> > > Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> > > Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> > > --- > > gcc/config/riscv/riscv-protos.h | 1 + > > gcc/config/riscv/riscv.cc | 14 ++++++++++++++ > > gcc/config/riscv/riscv.h | 4 ++++ > > .../gcc.target/riscv/rvv/xtheadvector/prefix.c | 12 ++++++++++++ > > 4 files changed, 31 insertions(+) > > create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c > > > > diff --git a/gcc/config/riscv/riscv-protos.h b/gcc/config/riscv/riscv-protos.h > > index 31049ef7523..5ea54b45703 100644 > > --- a/gcc/config/riscv/riscv-protos.h > > +++ b/gcc/config/riscv/riscv-protos.h > > @@ -102,6 +102,7 @@ struct riscv_address_info { > > }; > > > > /* Routines implemented in riscv.cc. */ > > +extern const char *riscv_asm_output_opcode (FILE *asm_out_file, const char *p); > > extern enum riscv_symbol_type riscv_classify_symbolic_expression (rtx); > > extern bool riscv_symbolic_constant_p (rtx, enum riscv_symbol_type *); > > extern int riscv_float_const_rtx_index_for_fli (rtx); > > diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc > > index 0d1cbc5cb5f..ea1d59d9cf2 100644 > > --- a/gcc/config/riscv/riscv.cc > > +++ b/gcc/config/riscv/riscv.cc > > @@ -5636,6 +5636,20 @@ riscv_get_v_regno_alignment (machine_mode mode) > > return lmul; > > } > > > > +/* Define ASM_OUTPUT_OPCODE to do anything special before > > + emitting an opcode. */ > > +const char * > > +riscv_asm_output_opcode (FILE *asm_out_file, const char *p) > > +{ > > + /* We need to add th. prefix to all the xtheadvector > > + insturctions here.*/ > > + if (TARGET_XTHEADVECTOR && current_output_insn != NULL_RTX && > > + p[0] == 'v') > > + fputs ("th.", asm_out_file); > > + > > + return p; > Just a formatting nit. The GNU standards break lines before the > operator, not after. So > if (TARGET_XTHEADVECTOR > && current_output_insn != NULL > && p[0] == 'v') > > Note that current_output_insn is "extern rtx_insn *", so use NULL, not > NULL_RTX. > > Neither of these nits require a new version for review. Just fix them. > > If Juzhe is fine with this, so am I. We can refine it if necessary later. > > jeff > > > > > ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. 2024-01-08 3:17 ` Re：Re: " joshua @ 2024-01-08 3:40 ` Kito Cheng 2024-01-10 3:01 ` Re：Re: " joshua 0 siblings, 1 reply; 133+ messages in thread From: Kito Cheng @ 2024-01-08 3:40 UTC (permalink / raw) To: joshua Cc: juzhe.zhong, jeffreyalaw, gcc-patches, Jim Wilson, palmer, andrew, philipp.tomsich, christoph.muellner, jinma, cooper.qu It depends on the timing when you send out the v1 patch to the mailing list, not the timing of when to merge, but of course it's case by case, I would say no IF it's still not ready when time is the end of Feb for this kind of big patch set. On Mon, Jan 8, 2024 at 11:17 AM joshua <cooper.joshua@linux.alibaba.com> wrote: > > Hi Kito, > > Thank you for your support. > So even during stage 4, we can merge this for GCC 14? > > > > > > ------------------------------------------------------------------ > 发件人：Kito Cheng <kito.cheng@gmail.com> > 发送时间：2024年1月8日(星期一) 11:06 > 收件人：joshua<cooper.joshua@linux.alibaba.com> > 抄 送："juzhe.zhong@rivai.ai"<juzhe.zhong@rivai.ai>; jeffreyalaw<jeffreyalaw@gmail.com>; "gcc-patches"<gcc-patches@gcc.gnu.org>; Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> > 主 题：Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. > > > I am ok with merging this for GCC 14, as we discussed several times in > the RISC-V GCC sync up meeting, I think at least we reach consensus > among Jeff Law, Palmer Dabbelt and me. > > But please be careful: don't break anything for standard vector stuff. > > On Mon, Jan 8, 2024 at 10:11 AM joshua <cooper.joshua@linux.alibaba.com> wrote: > > > > Hi Juzhe, > > > > Stage 3 will close today and there are still some patches that > > haven't been reviewed left. > > So is it possible to get xtheadvector merged in GCC-14? > > We emailed Kito regarding this, but haven't got any reply yet. > > > > Joshua > > > > > > > > > > > > > > ------------------------------------------------------------------ > > 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> > > 发送时间：2024年1月4日(星期四) 17:18 > > 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; jeffreyalaw<jeffreyalaw@gmail.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> > > 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> > > 主 题：Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. > > > > > > \ No newline at end of file > > Each file needs newline. > > > > > > I am not able to review arch stuff. This needs kito. > > > > > > Besides, Andrew Pinski want us defer theadvector to GCC-15. > > > > > > I have no strong opinion here. > > > > > > juzhe.zhong@rivai.ai > > > > > > 发件人： joshua > > 发送时间： 2024-01-04 17:15 > > 收件人： 钟居哲; Jeff Law; gcc-patches > > 抄送： jim.wilson.gcc; palmer; andrew; philipp.tomsich; Christoph Müllner; jinma; Cooper Qu > > 主题： Re：Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. > > > > Hi Juzhe, > > > > So is the following patch that this patch relies on OK to commit? > > https://gcc.gnu.org/pipermail/gcc-patches/2023-December/641533.html > > > > Joshua > > > > > > > > > > ------------------------------------------------------------------ > > 发件人：钟居哲 <juzhe.zhong@rivai.ai> > > 发送时间：2024年1月2日(星期二) 06:57 > > 收件人：Jeff Law<jeffreyalaw@gmail.com>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> > > 抄 送："jim.wilson.gcc"<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; "Christoph Müllner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; Cooper Qu<cooper.qu@linux.alibaba.com> > > 主 题：Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. > > > > > > This is Ok from my side. > > But before commit this patch, I think we need this patch first: > > https://gcc.gnu.org/pipermail/gcc-patches/2023-December/641533.html > > > > > > I will be back to work so I will take a look at other patches today. > > juzhe.zhong@rivai.ai > > > > > > From: Jeff Law > > Date: 2024-01-01 01:43 > > To: Jun Sha (Joshua); gcc-patches > > CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; christoph.muellner; juzhe.zhong; Jin Ma; Xianmiao Qu > > Subject: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. > > > > > > > > On 12/28/23 21:19, Jun Sha (Joshua) wrote: > > > This patch adds th. prefix to all XTheadVector instructions by > > > implementing new assembly output functions. We only check the > > > prefix is 'v', so that no extra attribute is needed. > > > > > > gcc/ChangeLog: > > > > > > * config/riscv/riscv-protos.h (riscv_asm_output_opcode): > > > New function to add assembler insn code prefix/suffix. > > > * config/riscv/riscv.cc (riscv_asm_output_opcode): Likewise. > > > * config/riscv/riscv.h (ASM_OUTPUT_OPCODE): Likewise. > > > > > > Co-authored-by: Jin Ma <jinma@linux.alibaba.com> > > > Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> > > > Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> > > > --- > > > gcc/config/riscv/riscv-protos.h | 1 + > > > gcc/config/riscv/riscv.cc | 14 ++++++++++++++ > > > gcc/config/riscv/riscv.h | 4 ++++ > > > .../gcc.target/riscv/rvv/xtheadvector/prefix.c | 12 ++++++++++++ > > > 4 files changed, 31 insertions(+) > > > create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c > > > > > > diff --git a/gcc/config/riscv/riscv-protos.h b/gcc/config/riscv/riscv-protos.h > > > index 31049ef7523..5ea54b45703 100644 > > > --- a/gcc/config/riscv/riscv-protos.h > > > +++ b/gcc/config/riscv/riscv-protos.h > > > @@ -102,6 +102,7 @@ struct riscv_address_info { > > > }; > > > > > > /* Routines implemented in riscv.cc. */ > > > +extern const char *riscv_asm_output_opcode (FILE *asm_out_file, const char *p); > > > extern enum riscv_symbol_type riscv_classify_symbolic_expression (rtx); > > > extern bool riscv_symbolic_constant_p (rtx, enum riscv_symbol_type *); > > > extern int riscv_float_const_rtx_index_for_fli (rtx); > > > diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc > > > index 0d1cbc5cb5f..ea1d59d9cf2 100644 > > > --- a/gcc/config/riscv/riscv.cc > > > +++ b/gcc/config/riscv/riscv.cc > > > @@ -5636,6 +5636,20 @@ riscv_get_v_regno_alignment (machine_mode mode) > > > return lmul; > > > } > > > > > > +/* Define ASM_OUTPUT_OPCODE to do anything special before > > > + emitting an opcode. */ > > > +const char * > > > +riscv_asm_output_opcode (FILE *asm_out_file, const char *p) > > > +{ > > > + /* We need to add th. prefix to all the xtheadvector > > > + insturctions here.*/ > > > + if (TARGET_XTHEADVECTOR && current_output_insn != NULL_RTX && > > > + p[0] == 'v') > > > + fputs ("th.", asm_out_file); > > > + > > > + return p; > > Just a formatting nit. The GNU standards break lines before the > > operator, not after. So > > if (TARGET_XTHEADVECTOR > > && current_output_insn != NULL > > && p[0] == 'v') > > > > Note that current_output_insn is "extern rtx_insn *", so use NULL, not > > NULL_RTX. > > > > Neither of these nits require a new version for review. Just fix them. > > > > If Juzhe is fine with this, so am I. We can refine it if necessary later. > > > > jeff > > > > > > > > > > ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re：Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. 2024-01-08 3:40 ` Kito Cheng @ 2024-01-10 3:01 ` joshua 0 siblings, 0 replies; 133+ messages in thread From: joshua @ 2024-01-10 3:01 UTC (permalink / raw) To: Kito Cheng Cc: juzhe.zhong, jeffreyalaw, gcc-patches, Jim Wilson, palmer, andrew, philipp.tomsich, christoph.muellner, jinma, cooper.qu Hi Kito, Thank you for your support again. I believe we can get all our xtheadvector patches ready before the end of Feb. May I please ping the arch patch again? https://gcc.gnu.org/pipermail/gcc-patches/2024-January/641801.html This is the patch that all the following patches rely on. Joshua ------------------------------------------------------------------ 发件人：Kito Cheng <kito.cheng@gmail.com> 发送时间：2024年1月8日(星期一) 11:40 收件人：joshua<cooper.joshua@linux.alibaba.com> 抄 送："juzhe.zhong@rivai.ai"<juzhe.zhong@rivai.ai>; jeffreyalaw<jeffreyalaw@gmail.com>; "gcc-patches"<gcc-patches@gcc.gnu.org>; Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. It depends on the timing when you send out the v1 patch to the mailing list, not the timing of when to merge, but of course it's case by case, I would say no IF it's still not ready when time is the end of Feb for this kind of big patch set. On Mon, Jan 8, 2024 at 11:17 AM joshua <cooper.joshua@linux.alibaba.com> wrote: > > Hi Kito, > > Thank you for your support. > So even during stage 4, we can merge this for GCC 14? > > > > > > ------------------------------------------------------------------ > 发件人：Kito Cheng <kito.cheng@gmail.com> > 发送时间：2024年1月8日(星期一) 11:06 > 收件人：joshua<cooper.joshua@linux.alibaba.com> > 抄 送："juzhe.zhong@rivai.ai"<juzhe.zhong@rivai.ai>; jeffreyalaw<jeffreyalaw@gmail.com>; "gcc-patches"<gcc-patches@gcc.gnu.org>; Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> > 主 题：Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. > > > I am ok with merging this for GCC 14, as we discussed several times in > the RISC-V GCC sync up meeting, I think at least we reach consensus > among Jeff Law, Palmer Dabbelt and me. > > But please be careful: don't break anything for standard vector stuff. > > On Mon, Jan 8, 2024 at 10:11 AM joshua <cooper.joshua@linux.alibaba.com> wrote: > > > > Hi Juzhe, > > > > Stage 3 will close today and there are still some patches that > > haven't been reviewed left. > > So is it possible to get xtheadvector merged in GCC-14? > > We emailed Kito regarding this, but haven't got any reply yet. > > > > Joshua > > > > > > > > > > > > > > ------------------------------------------------------------------ > > 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> > > 发送时间：2024年1月4日(星期四) 17:18 > > 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; jeffreyalaw<jeffreyalaw@gmail.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> > > 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> > > 主 题：Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. > > > > > > \ No newline at end of file > > Each file needs newline. > > > > > > I am not able to review arch stuff. This needs kito. > > > > > > Besides, Andrew Pinski want us defer theadvector to GCC-15. > > > > > > I have no strong opinion here. > > > > > > juzhe.zhong@rivai.ai > > > > > > 发件人： joshua > > 发送时间： 2024-01-04 17:15 > > 收件人： 钟居哲; Jeff Law; gcc-patches > > 抄送： jim.wilson.gcc; palmer; andrew; philipp.tomsich; Christoph Müllner; jinma; Cooper Qu > > 主题： Re：Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. > > > > Hi Juzhe, > > > > So is the following patch that this patch relies on OK to commit? > > https://gcc.gnu.org/pipermail/gcc-patches/2023-December/641533.html > > > > Joshua > > > > > > > > > > ------------------------------------------------------------------ > > 发件人：钟居哲 <juzhe.zhong@rivai.ai> > > 发送时间：2024年1月2日(星期二) 06:57 > > 收件人：Jeff Law<jeffreyalaw@gmail.com>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> > > 抄 送："jim.wilson.gcc"<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; "Christoph Müllner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; Cooper Qu<cooper.qu@linux.alibaba.com> > > 主 题：Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. > > > > > > This is Ok from my side. > > But before commit this patch, I think we need this patch first: > > https://gcc.gnu.org/pipermail/gcc-patches/2023-December/641533.html > > > > > > I will be back to work so I will take a look at other patches today. > > juzhe.zhong@rivai.ai > > > > > > From: Jeff Law > > Date: 2024-01-01 01:43 > > To: Jun Sha (Joshua); gcc-patches > > CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; christoph.muellner; juzhe.zhong; Jin Ma; Xianmiao Qu > > Subject: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. > > > > > > > > On 12/28/23 21:19, Jun Sha (Joshua) wrote: > > > This patch adds th. prefix to all XTheadVector instructions by > > > implementing new assembly output functions. We only check the > > > prefix is 'v', so that no extra attribute is needed. > > > > > > gcc/ChangeLog: > > > > > > * config/riscv/riscv-protos.h (riscv_asm_output_opcode): > > > New function to add assembler insn code prefix/suffix. > > > * config/riscv/riscv.cc (riscv_asm_output_opcode): Likewise. > > > * config/riscv/riscv.h (ASM_OUTPUT_OPCODE): Likewise. > > > > > > Co-authored-by: Jin Ma <jinma@linux.alibaba.com> > > > Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> > > > Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> > > > --- > > > gcc/config/riscv/riscv-protos.h | 1 + > > > gcc/config/riscv/riscv.cc | 14 ++++++++++++++ > > > gcc/config/riscv/riscv.h | 4 ++++ > > > .../gcc.target/riscv/rvv/xtheadvector/prefix.c | 12 ++++++++++++ > > > 4 files changed, 31 insertions(+) > > > create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c > > > > > > diff --git a/gcc/config/riscv/riscv-protos.h b/gcc/config/riscv/riscv-protos.h > > > index 31049ef7523..5ea54b45703 100644 > > > --- a/gcc/config/riscv/riscv-protos.h > > > +++ b/gcc/config/riscv/riscv-protos.h > > > @@ -102,6 +102,7 @@ struct riscv_address_info { > > > }; > > > > > > /* Routines implemented in riscv.cc. */ > > > +extern const char *riscv_asm_output_opcode (FILE *asm_out_file, const char *p); > > > extern enum riscv_symbol_type riscv_classify_symbolic_expression (rtx); > > > extern bool riscv_symbolic_constant_p (rtx, enum riscv_symbol_type *); > > > extern int riscv_float_const_rtx_index_for_fli (rtx); > > > diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc > > > index 0d1cbc5cb5f..ea1d59d9cf2 100644 > > > --- a/gcc/config/riscv/riscv.cc > > > +++ b/gcc/config/riscv/riscv.cc > > > @@ -5636,6 +5636,20 @@ riscv_get_v_regno_alignment (machine_mode mode) > > > return lmul; > > > } > > > > > > +/* Define ASM_OUTPUT_OPCODE to do anything special before > > > + emitting an opcode. */ > > > +const char * > > > +riscv_asm_output_opcode (FILE *asm_out_file, const char *p) > > > +{ > > > + /* We need to add th. prefix to all the xtheadvector > > > + insturctions here.*/ > > > + if (TARGET_XTHEADVECTOR && current_output_insn != NULL_RTX && > > > + p[0] == 'v') > > > + fputs ("th.", asm_out_file); > > > + > > > + return p; > > Just a formatting nit. The GNU standards break lines before the > > operator, not after. So > > if (TARGET_XTHEADVECTOR > > && current_output_insn != NULL > > && p[0] == 'v') > > > > Note that current_output_insn is "extern rtx_insn *", so use NULL, not > > NULL_RTX. > > > > Neither of these nits require a new version for review. Just fix them. > > > > If Juzhe is fine with this, so am I. We can refine it if necessary later. > > > > jeff > > > > > > > > > > ^ permalink raw reply [flat|nested] 133+ messages in thread
* [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. 2023-12-29 4:19 ` [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector Jun Sha (Joshua) 2023-12-31 17:43 ` Jeff Law @ 2024-01-03 6:08 ` Jun Sha (Joshua) 2024-01-08 23:04 ` 钟居哲 1 sibling, 1 reply; 133+ messages in thread From: Jun Sha (Joshua) @ 2024-01-03 6:08 UTC (permalink / raw) To: gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, Jun Sha (Joshua), Jin Ma, Xianmiao Qu This patch adds th. prefix to all XTheadVector instructions by implementing new assembly output functions. We only check the prefix is 'v', so that no extra attribute is needed. gcc/ChangeLog: * config/riscv/riscv-protos.h (riscv_asm_output_opcode): New function to add assembler insn code prefix/suffix. (th_asm_output_opcode): Thead function to add assembler insn code prefix/suffix. * config/riscv/riscv.cc (riscv_asm_output_opcode): Likewise * config/riscv/riscv.h (ASM_OUTPUT_OPCODE): Likewise. * config/riscv/thead.cc (th_asm_output_opcode): Likewise gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/prefix.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config/riscv/riscv-protos.h | 2 ++ gcc/config/riscv/riscv.cc | 11 +++++++++++ gcc/config/riscv/riscv.h | 4 ++++ gcc/config/riscv/thead.cc | 13 +++++++++++++ .../gcc.target/riscv/rvv/xtheadvector/prefix.c | 12 ++++++++++++ 5 files changed, 42 insertions(+) create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c diff --git a/gcc/config/riscv/riscv-protos.h b/gcc/config/riscv/riscv-protos.h index 31049ef7523..71724dabdb5 100644 --- a/gcc/config/riscv/riscv-protos.h +++ b/gcc/config/riscv/riscv-protos.h @@ -102,6 +102,7 @@ struct riscv_address_info { }; /* Routines implemented in riscv.cc. */ +extern const char *riscv_asm_output_opcode (FILE *asm_out_file, const char *p); extern enum riscv_symbol_type riscv_classify_symbolic_expression (rtx); extern bool riscv_symbolic_constant_p (rtx, enum riscv_symbol_type *); extern int riscv_float_const_rtx_index_for_fli (rtx); @@ -717,6 +718,7 @@ extern void th_mempair_prepare_save_restore_operands (rtx[4], bool, int, HOST_WIDE_INT, int, HOST_WIDE_INT); extern void th_mempair_save_restore_regs (rtx[4], bool, machine_mode); +extern const char *th_asm_output_opcode (FILE *asm_out_file, const char *p); #ifdef RTX_CODE extern const char* th_mempair_output_move (rtx[4], bool, machine_mode, RTX_CODE); diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index 0d1cbc5cb5f..51878797287 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -5636,6 +5636,17 @@ riscv_get_v_regno_alignment (machine_mode mode) return lmul; } +/* Define ASM_OUTPUT_OPCODE to do anything special before + emitting an opcode. */ +const char * +riscv_asm_output_opcode (FILE *asm_out_file, const char *p) +{ + if (TARGET_XTHEADVECTOR) + return th_asm_output_opcode (asm_out_file, p); + + return p; +} + /* Implement TARGET_PRINT_OPERAND. The RISCV-specific operand codes are: 'h' Print the high-part relocation associated with OP, after stripping diff --git a/gcc/config/riscv/riscv.h b/gcc/config/riscv/riscv.h index 6df9ec73c5e..c33361a254d 100644 --- a/gcc/config/riscv/riscv.h +++ b/gcc/config/riscv/riscv.h @@ -826,6 +826,10 @@ extern enum riscv_cc get_riscv_cc (const rtx use); asm_fprintf ((FILE), "%U%s", (NAME)); \ } while (0) +#undef ASM_OUTPUT_OPCODE +#define ASM_OUTPUT_OPCODE(STREAM, PTR) \ + (PTR) = riscv_asm_output_opcode(STREAM, PTR) + #define JUMP_TABLES_IN_TEXT_SECTION 0 #define CASE_VECTOR_MODE SImode #define CASE_VECTOR_PC_RELATIVE (riscv_cmodel != CM_MEDLOW) diff --git a/gcc/config/riscv/thead.cc b/gcc/config/riscv/thead.cc index 20353995931..dc3aed3904d 100644 --- a/gcc/config/riscv/thead.cc +++ b/gcc/config/riscv/thead.cc @@ -883,6 +883,19 @@ th_output_move (rtx dest, rtx src) return NULL; } +/* Define ASM_OUTPUT_OPCODE to do anything special before + emitting an opcode. */ +const char * +th_asm_output_opcode (FILE *asm_out_file, const char *p) +{ + /* We need to add th. prefix to all the xtheadvector + instructions here.*/ + if (current_output_insn != NULL && p[0] == 'v') + fputs ("th.", asm_out_file); + + return p; +} + /* Implement TARGET_PRINT_OPERAND_ADDRESS for XTheadMemIdx. */ bool diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c new file mode 100644 index 00000000000..eee727ef6b4 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c @@ -0,0 +1,12 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gc_xtheadvector -mabi=ilp32 -O0" } */ + +#include "riscv_vector.h" + +vint32m1_t +prefix (vint32m1_t vx, vint32m1_t vy, size_t vl) +{ + return __riscv_vadd_vv_i32m1 (vx, vy, vl); +} + +/* { dg-final { scan-assembler {\mth\.v\M} } } */ -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. 2024-01-03 6:08 ` Jun Sha (Joshua) @ 2024-01-08 23:04 ` 钟居哲 2024-01-09 17:49 ` Jeff Law 0 siblings, 1 reply; 133+ messages in thread From: 钟居哲 @ 2024-01-08 23:04 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, Jeff Law, Christoph Müllner, cooper.joshua, jinma, Cooper Qu [-- Attachment #1: Type: text/plain, Size: 5164 bytes --] This patch looks ok from myside. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-03 14:08 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. This patch adds th. prefix to all XTheadVector instructions by implementing new assembly output functions. We only check the prefix is 'v', so that no extra attribute is needed. gcc/ChangeLog: * config/riscv/riscv-protos.h (riscv_asm_output_opcode): New function to add assembler insn code prefix/suffix. (th_asm_output_opcode): Thead function to add assembler insn code prefix/suffix. * config/riscv/riscv.cc (riscv_asm_output_opcode): Likewise * config/riscv/riscv.h (ASM_OUTPUT_OPCODE): Likewise. * config/riscv/thead.cc (th_asm_output_opcode): Likewise gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/prefix.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config/riscv/riscv-protos.h | 2 ++ gcc/config/riscv/riscv.cc | 11 +++++++++++ gcc/config/riscv/riscv.h | 4 ++++ gcc/config/riscv/thead.cc | 13 +++++++++++++ .../gcc.target/riscv/rvv/xtheadvector/prefix.c | 12 ++++++++++++ 5 files changed, 42 insertions(+) create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c diff --git a/gcc/config/riscv/riscv-protos.h b/gcc/config/riscv/riscv-protos.h index 31049ef7523..71724dabdb5 100644 --- a/gcc/config/riscv/riscv-protos.h +++ b/gcc/config/riscv/riscv-protos.h @@ -102,6 +102,7 @@ struct riscv_address_info { }; /* Routines implemented in riscv.cc. */ +extern const char *riscv_asm_output_opcode (FILE *asm_out_file, const char *p); extern enum riscv_symbol_type riscv_classify_symbolic_expression (rtx); extern bool riscv_symbolic_constant_p (rtx, enum riscv_symbol_type *); extern int riscv_float_const_rtx_index_for_fli (rtx); @@ -717,6 +718,7 @@ extern void th_mempair_prepare_save_restore_operands (rtx[4], bool, int, HOST_WIDE_INT, int, HOST_WIDE_INT); extern void th_mempair_save_restore_regs (rtx[4], bool, machine_mode); +extern const char *th_asm_output_opcode (FILE *asm_out_file, const char *p); #ifdef RTX_CODE extern const char* th_mempair_output_move (rtx[4], bool, machine_mode, RTX_CODE); diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index 0d1cbc5cb5f..51878797287 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -5636,6 +5636,17 @@ riscv_get_v_regno_alignment (machine_mode mode) return lmul; } +/* Define ASM_OUTPUT_OPCODE to do anything special before + emitting an opcode. */ +const char * +riscv_asm_output_opcode (FILE *asm_out_file, const char *p) +{ + if (TARGET_XTHEADVECTOR) + return th_asm_output_opcode (asm_out_file, p); + + return p; +} + /* Implement TARGET_PRINT_OPERAND. The RISCV-specific operand codes are: 'h' Print the high-part relocation associated with OP, after stripping diff --git a/gcc/config/riscv/riscv.h b/gcc/config/riscv/riscv.h index 6df9ec73c5e..c33361a254d 100644 --- a/gcc/config/riscv/riscv.h +++ b/gcc/config/riscv/riscv.h @@ -826,6 +826,10 @@ extern enum riscv_cc get_riscv_cc (const rtx use); asm_fprintf ((FILE), "%U%s", (NAME)); \ } while (0) +#undef ASM_OUTPUT_OPCODE +#define ASM_OUTPUT_OPCODE(STREAM, PTR) \ + (PTR) = riscv_asm_output_opcode(STREAM, PTR) + #define JUMP_TABLES_IN_TEXT_SECTION 0 #define CASE_VECTOR_MODE SImode #define CASE_VECTOR_PC_RELATIVE (riscv_cmodel != CM_MEDLOW) diff --git a/gcc/config/riscv/thead.cc b/gcc/config/riscv/thead.cc index 20353995931..dc3aed3904d 100644 --- a/gcc/config/riscv/thead.cc +++ b/gcc/config/riscv/thead.cc @@ -883,6 +883,19 @@ th_output_move (rtx dest, rtx src) return NULL; } +/* Define ASM_OUTPUT_OPCODE to do anything special before + emitting an opcode. */ +const char * +th_asm_output_opcode (FILE *asm_out_file, const char *p) +{ + /* We need to add th. prefix to all the xtheadvector + instructions here.*/ + if (current_output_insn != NULL && p[0] == 'v') + fputs ("th.", asm_out_file); + + return p; +} + /* Implement TARGET_PRINT_OPERAND_ADDRESS for XTheadMemIdx. */ bool diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c new file mode 100644 index 00000000000..eee727ef6b4 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c @@ -0,0 +1,12 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gc_xtheadvector -mabi=ilp32 -O0" } */ + +#include "riscv_vector.h" + +vint32m1_t +prefix (vint32m1_t vx, vint32m1_t vy, size_t vl) +{ + return __riscv_vadd_vv_i32m1 (vx, vy, vl); +} + +/* { dg-final { scan-assembler {\mth\.v\M} } } */ -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. 2024-01-08 23:04 ` 钟居哲 @ 2024-01-09 17:49 ` Jeff Law 2024-01-09 22:35 ` 钟居哲 0 siblings, 1 reply; 133+ messages in thread From: Jeff Law @ 2024-01-09 17:49 UTC (permalink / raw) To: 钟居哲, cooper.joshua, gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, Christoph Müllner, jinma, Cooper Qu On 1/8/24 16:04, 钟居哲 wrote: > This patch looks ok from myside. Likewise. So I think the only question for this specific patch is whether or not it makes sense to include it now or wait for more of the thead bits to get to acceptance. I tend to think it should wait since I don't think it has any value without the rest of the thead vector changes and it's not 100% clear if those changes are going to make it into gcc-14 or not. Jeff ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. 2024-01-09 17:49 ` Jeff Law @ 2024-01-09 22:35 ` 钟居哲 0 siblings, 0 replies; 133+ messages in thread From: 钟居哲 @ 2024-01-09 22:35 UTC (permalink / raw) To: Jeff Law, cooper.joshua, gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, Christoph Müllner, jinma, Cooper Qu [-- Attachment #1: Type: text/plain, Size: 862 bytes --] Yes. I aggree with you that we should wait until all theadvector are acccepted. Thanks. juzhe.zhong@rivai.ai From: Jeff Law Date: 2024-01-10 01:49 To: 钟居哲; cooper.joshua; gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; Christoph Müllner; jinma; Cooper Qu Subject: Re: [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. On 1/8/24 16:04, 钟居哲 wrote: > This patch looks ok from myside. Likewise. So I think the only question for this specific patch is whether or not it makes sense to include it now or wait for more of the thead bits to get to acceptance. I tend to think it should wait since I don't think it has any value without the rest of the thead vector changes and it's not 100% clear if those changes are going to make it into gcc-14 or not. Jeff ^ permalink raw reply [flat|nested] 133+ messages in thread
* [PATCH v4] RISC-V: Handle differences between XTheadvector and Vector 2023-12-29 4:03 [PATCH v4] RISC-V: Support XTheadVector extension Jun Sha (Joshua) ` (3 preceding siblings ...) 2023-12-29 4:19 ` [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector Jun Sha (Joshua) @ 2023-12-29 4:21 ` Jun Sha (Joshua) 2024-01-02 2:00 ` juzhe.zhong 2024-01-02 12:39 ` [PATCH " Jun Sha (Joshua) 2023-12-29 4:21 ` [PATCH v4 6/6] RISC-V: Add support for xtheadvector-specific intrinsics Jun Sha (Joshua) ` (3 subsequent siblings) 8 siblings, 2 replies; 133+ messages in thread From: Jun Sha (Joshua) @ 2023-12-29 4:21 UTC (permalink / raw) To: gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, Jun Sha (Joshua), Jin Ma, Xianmiao Qu This patch is to handle the differences in instruction generation between Vector and XTheadVector. In this version, we only support partial xtheadvector instructions that leverage directly from current RVV1.0 with simple adding "th." prefix. For different name xtheadvector instructions but share same patterns as RVV1.0 instructions, we will use ASM targethook to rewrite the whole string of the instructions in the following patches. For some vector patterns that cannot be avoided, we use "!TARGET_XTHEADVECTOR" to disable them in vector.md in order not to generate instructions that xtheadvector does not support, like vmv1r and vsext.vf2. gcc/ChangeLog: * config.gcc: Add files for XTheadVector intrinsics. * config/riscv/autovec.md: Guard XTheadVector. * config/riscv/riscv-string.cc (expand_block_move): Guard XTheadVector. * config/riscv/riscv-v.cc (legitimize_move): New expansion. (get_prefer_tail_policy): Give specific value for tail. (get_prefer_mask_policy): Give specific value for mask. (vls_mode_valid_p): Avoid autovec. * config/riscv/riscv-vector-builtins-shapes.cc (check_type): (build_one): New function. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): (DEF_THEAD_RVV_FUNCTION): Add new marcos. (check_required_extensions): (handle_pragma_vector): * config/riscv/riscv-vector-builtins.h (RVV_REQUIRE_VECTOR): (RVV_REQUIRE_XTHEADVECTOR): Add RVV_REQUIRE_VECTOR and RVV_REQUIRE_XTHEADVECTOR. (struct function_group_info): * config/riscv/riscv-vector-switch.def (ENTRY): Disable fractional mode for the XTheadVector extension. (TUPLE_ENTRY): Likewise. * config/riscv/riscv-vsetvl.cc: Add functions for xtheadvector. * config/riscv/riscv.cc (riscv_v_ext_vls_mode_p): Guard XTheadVector. (riscv_v_adjust_bytesize): Likewise. (riscv_preferred_simd_mode): Likewsie. (riscv_autovectorize_vector_modes): Likewise. (riscv_vector_mode_supported_any_target_p): Likewise. (TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P): Likewise. * config/riscv/vector-iterators.md: Remove fractional LMUL. * config/riscv/vector.md: Include thead-vector.md. * config/riscv/riscv_th_vector.h: New file. * config/riscv/thead-vector.md: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/base/pragma-1.c: Add XTheadVector. * gcc.target/riscv/rvv/base/abi-1.c: Exclude XTheadVector. * lib/target-supports.exp: Add target for XTheadVector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config.gcc | 2 +- gcc/config/riscv/autovec.md | 2 +- gcc/config/riscv/predicates.md | 3 +- gcc/config/riscv/riscv-string.cc | 3 + gcc/config/riscv/riscv-v.cc | 13 +- .../riscv/riscv-vector-builtins-bases.cc | 3 + .../riscv/riscv-vector-builtins-shapes.cc | 23 +++ gcc/config/riscv/riscv-vector-switch.def | 150 +++++++------- gcc/config/riscv/riscv-vsetvl.cc | 10 + gcc/config/riscv/riscv.cc | 20 +- gcc/config/riscv/riscv_th_vector.h | 49 +++++ gcc/config/riscv/thead-vector.md | 142 +++++++++++++ gcc/config/riscv/vector-iterators.md | 186 +++++++++--------- gcc/config/riscv/vector.md | 36 +++- .../gcc.target/riscv/rvv/base/abi-1.c | 2 +- .../gcc.target/riscv/rvv/base/pragma-1.c | 2 +- gcc/testsuite/lib/target-supports.exp | 12 ++ 17 files changed, 471 insertions(+), 187 deletions(-) create mode 100644 gcc/config/riscv/riscv_th_vector.h create mode 100644 gcc/config/riscv/thead-vector.md diff --git a/gcc/config.gcc b/gcc/config.gcc index f0676c830e8..1445d98c147 100644 --- a/gcc/config.gcc +++ b/gcc/config.gcc @@ -549,7 +549,7 @@ riscv*) extra_objs="${extra_objs} riscv-vector-builtins.o riscv-vector-builtins-shapes.o riscv-vector-builtins-bases.o" extra_objs="${extra_objs} thead.o riscv-target-attr.o" d_target_objs="riscv-d.o" - extra_headers="riscv_vector.h" + extra_headers="riscv_vector.h riscv_th_vector.h" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.cc" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.h" ;; diff --git a/gcc/config/riscv/autovec.md b/gcc/config/riscv/autovec.md index 8b8a92f10a1..1fac56c7095 100644 --- a/gcc/config/riscv/autovec.md +++ b/gcc/config/riscv/autovec.md @@ -2579,7 +2579,7 @@ [(match_operand 0 "register_operand") (match_operand 1 "memory_operand") (match_operand:ANYI 2 "const_int_operand")] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" { riscv_vector::expand_rawmemchr(<MODE>mode, operands[0], operands[1], operands[2]); diff --git a/gcc/config/riscv/predicates.md b/gcc/config/riscv/predicates.md index b87a6900841..c5ab86d5db9 100644 --- a/gcc/config/riscv/predicates.md +++ b/gcc/config/riscv/predicates.md @@ -428,7 +428,8 @@ ;; Predicates for the V extension. (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (match_test "!TARGET_XTHEADVECTOR || rtx_equal_p (op, const0_rtx)") + (match_operand 0 "const_csr_operand")))) (define_special_predicate "autovec_length_operand" (ior (match_operand 0 "pmode_register_operand") diff --git a/gcc/config/riscv/riscv-string.cc b/gcc/config/riscv/riscv-string.cc index 11c1f74d0b3..ec8f3486fd8 100644 --- a/gcc/config/riscv/riscv-string.cc +++ b/gcc/config/riscv/riscv-string.cc @@ -808,6 +808,9 @@ expand_block_move (rtx dst_in, rtx src_in, rtx length_in) bnez a2, loop # Any more? ret # Return */ + if (TARGET_XTHEADVECTOR) + return false; + gcc_assert (TARGET_VECTOR); HOST_WIDE_INT potential_ew diff --git a/gcc/config/riscv/riscv-v.cc b/gcc/config/riscv/riscv-v.cc index 038ab084a37..5e9e45aecd2 100644 --- a/gcc/config/riscv/riscv-v.cc +++ b/gcc/config/riscv/riscv-v.cc @@ -1523,6 +1523,13 @@ legitimize_move (rtx dest, rtx *srcp) return true; } + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (mode, dest, src, + RVV_VLMAX, GEN_INT(VLMAX))); + return true; + } + if (riscv_v_ext_vls_mode_p (mode)) { if (GET_MODE_NUNITS (mode).to_constant () <= 31) @@ -1772,7 +1779,7 @@ get_prefer_tail_policy () compiler pick up either agnostic or undisturbed. Maybe we will have a compile option like -mprefer=agnostic to set this value???. */ - return TAIL_ANY; + return TARGET_XTHEADVECTOR ? TAIL_AGNOSTIC : TAIL_ANY; } /* Get prefer mask policy. */ @@ -1783,7 +1790,7 @@ get_prefer_mask_policy () compiler pick up either agnostic or undisturbed. Maybe we will have a compile option like -mprefer=agnostic to set this value???. */ - return MASK_ANY; + return TARGET_XTHEADVECTOR ? MASK_UNDISTURBED : MASK_ANY; } /* Get avl_type rtx. */ @@ -4383,7 +4390,7 @@ cmp_lmul_gt_one (machine_mode mode) bool vls_mode_valid_p (machine_mode vls_mode) { - if (!TARGET_VECTOR) + if (!TARGET_VECTOR || TARGET_XTHEADVECTOR) return false; if (riscv_autovec_preference == RVV_SCALABLE) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index c51affde353..2918c07ebf3 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -133,6 +133,9 @@ public: = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + if (TARGET_XTHEADVECTOR) + return e.generate_insn (code_for_th_vsetvl_no_side_effects (Pmode)); + /* TAIL_ANY. */ e.add_input_operand (Pmode, gen_int_mode (get_prefer_tail_policy (), Pmode)); diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 4a754e0228f..6b49404a1fa 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -33,6 +33,25 @@ namespace riscv_vector { +/* Check whether the RETURN_TYPE and ARGUMENT_TYPES are + valid for the function. */ + +static bool +check_type (tree return_type, vec<tree> &argument_types) +{ + tree arg; + unsigned i; + + if (!return_type) + return false; + + FOR_EACH_VEC_ELT (argument_types, i, arg) + if (!arg) + return false; + + return true; +} + /* Add one function instance for GROUP, using operand suffix at index OI, mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ static void @@ -49,6 +68,10 @@ build_one (function_builder &b, const function_group_info &group, group.ops_infos.types[vec_type_idx].index); b.allocate_argument_types (function_instance, argument_types); b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + b.add_overloaded_function (function_instance, *group.shape); b.add_unique_function (function_instance, (*group.shape), return_type, argument_types); diff --git a/gcc/config/riscv/riscv-vector-switch.def b/gcc/config/riscv/riscv-vector-switch.def index 5c9f9bcbc3e..f7a66b34bae 100644 --- a/gcc/config/riscv/riscv-vector-switch.def +++ b/gcc/config/riscv/riscv-vector-switch.def @@ -68,9 +68,9 @@ Encode the ratio of SEW/LMUL into the mask types. #endif /* Disable modes if TARGET_MIN_VLEN == 32. */ -ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) -ENTRY (RVVMF32BI, true, LMUL_F4, 32) -ENTRY (RVVMF16BI, true, LMUL_F2, 16) +ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F8, 64) +ENTRY (RVVMF32BI, true, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F4, 32) +ENTRY (RVVMF16BI, true, TARGET_XTHEADVECTOR ? LMUL_1 : LMUL_F2 , 16) ENTRY (RVVMF8BI, true, LMUL_1, 8) ENTRY (RVVMF4BI, true, LMUL_2, 4) ENTRY (RVVMF2BI, true, LMUL_4, 2) @@ -81,39 +81,39 @@ ENTRY (RVVM8QI, true, LMUL_8, 1) ENTRY (RVVM4QI, true, LMUL_4, 2) ENTRY (RVVM2QI, true, LMUL_2, 4) ENTRY (RVVM1QI, true, LMUL_1, 8) -ENTRY (RVVMF2QI, true, LMUL_F2, 16) -ENTRY (RVVMF4QI, true, LMUL_F4, 32) -ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) +ENTRY (RVVMF2QI, !TARGET_XTHEADVECTOR, LMUL_F2, 16) +ENTRY (RVVMF4QI, !TARGET_XTHEADVECTOR, LMUL_F4, 32) +ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F8, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8HI, true, LMUL_8, 2) ENTRY (RVVM4HI, true, LMUL_4, 4) ENTRY (RVVM2HI, true, LMUL_2, 8) ENTRY (RVVM1HI, true, LMUL_1, 16) -ENTRY (RVVMF2HI, true, LMUL_F2, 32) -ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HI, !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_16. */ ENTRY (RVVM8HF, TARGET_VECTOR_ELEN_FP_16, LMUL_8, 2) ENTRY (RVVM4HF, TARGET_VECTOR_ELEN_FP_16, LMUL_4, 4) ENTRY (RVVM2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_2, 8) ENTRY (RVVM1HF, TARGET_VECTOR_ELEN_FP_16, LMUL_1, 16) -ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_F2, 32) -ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8SI, true, LMUL_8, 4) ENTRY (RVVM4SI, true, LMUL_4, 8) ENTRY (RVVM2SI, true, LMUL_2, 16) ENTRY (RVVM1SI, true, LMUL_1, 32) -ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_32. */ ENTRY (RVVM8SF, TARGET_VECTOR_ELEN_FP_32, LMUL_8, 4) ENTRY (RVVM4SF, TARGET_VECTOR_ELEN_FP_32, LMUL_4, 8) ENTRY (RVVM2SF, TARGET_VECTOR_ELEN_FP_32, LMUL_2, 16) ENTRY (RVVM1SF, TARGET_VECTOR_ELEN_FP_32, LMUL_1, 32) -ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if !TARGET_VECTOR_ELEN_64. */ ENTRY (RVVM8DI, TARGET_VECTOR_ELEN_64, LMUL_8, 8) @@ -140,127 +140,127 @@ ENTRY (RVVM1DF, TARGET_VECTOR_ELEN_FP_64, LMUL_1, 64) #endif TUPLE_ENTRY (RVVM1x8QI, true, RVVM1QI, 8, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x8QI, true, RVVMF2QI, 8, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x8QI, true, RVVMF4QI, 8, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 8, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x8QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 8, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x8QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 8, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 8, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x7QI, true, RVVM1QI, 7, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x7QI, true, RVVMF2QI, 7, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x7QI, true, RVVMF4QI, 7, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 7, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x7QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 7, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x7QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 7, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 7, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x6QI, true, RVVM1QI, 6, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x6QI, true, RVVMF2QI, 6, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x6QI, true, RVVMF4QI, 6, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 6, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x6QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 6, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x6QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 6, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 6, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x5QI, true, RVVM1QI, 5, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x5QI, true, RVVMF2QI, 5, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x5QI, true, RVVMF4QI, 5, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 5, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x5QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 5, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x5QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 5, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 5, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x4QI, true, RVVM2QI, 4, LMUL_2, 4) TUPLE_ENTRY (RVVM1x4QI, true, RVVM1QI, 4, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x4QI, true, RVVMF2QI, 4, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x4QI, true, RVVMF4QI, 4, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 4, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x4QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 4, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x4QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 4, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 4, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x3QI, true, RVVM2QI, 3, LMUL_2, 4) TUPLE_ENTRY (RVVM1x3QI, true, RVVM1QI, 3, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x3QI, true, RVVMF2QI, 3, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x3QI, true, RVVMF4QI, 3, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 3, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x3QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 3, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x3QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 3, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 3, LMUL_F8, 64) TUPLE_ENTRY (RVVM4x2QI, true, RVVM4QI, 2, LMUL_4, 2) TUPLE_ENTRY (RVVM2x2QI, true, RVVM2QI, 2, LMUL_2, 4) TUPLE_ENTRY (RVVM1x2QI, true, RVVM1QI, 2, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x2QI, true, RVVMF2QI, 2, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x2QI, true, RVVMF4QI, 2, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 2, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x2QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 2, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x2QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 2, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 2, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x8HI, true, RVVM1HI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HI, true, RVVMF2HI, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HI, true, RVVM1HI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HI, true, RVVMF2HI, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HI, true, RVVM1HI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HI, true, RVVMF2HI, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HI, true, RVVM1HI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HI, true, RVVMF2HI, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HI, true, RVVM2HI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HI, true, RVVM1HI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HI, true, RVVMF2HI, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HI, true, RVVM2HI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HI, true, RVVM1HI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HI, true, RVVMF2HI, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HI, true, RVVM4HI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HI, true, RVVM2HI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HI, true, RVVM1HI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HI, true, RVVMF2HI, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM4HF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8SI, true, RVVM1SI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SI, true, RVVM1SI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SI, true, RVVM1SI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SI, true, RVVM1SI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SI, true, RVVM2SI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SI, true, RVVM1SI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SI, true, RVVM2SI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SI, true, RVVM1SI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SI, true, RVVM4SI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SI, true, RVVM2SI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SI, true, RVVM1SI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM4SF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 8, LMUL_1, 16) TUPLE_ENTRY (RVVM1x7DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 7, LMUL_1, 16) diff --git a/gcc/config/riscv/riscv-vsetvl.cc b/gcc/config/riscv/riscv-vsetvl.cc index eabaef80f89..c726253c107 100644 --- a/gcc/config/riscv/riscv-vsetvl.cc +++ b/gcc/config/riscv/riscv-vsetvl.cc @@ -1117,6 +1117,16 @@ public: avl = GEN_INT (0); rtx sew = gen_int_mode (get_sew (), Pmode); rtx vlmul = gen_int_mode (get_vlmul (), Pmode); + + if (TARGET_XTHEADVECTOR) { + if (change_vtype_only_p ()) + return gen_th_vsetvl_vtype_change_only (sew, vlmul); + else if (has_vl () && !ignore_vl) + return gen_th_vsetvl (Pmode, get_vl (), avl, sew, vlmul); + else + return gen_th_vsetvl_discard_result (Pmode, avl, sew, vlmul); + } + rtx ta = gen_int_mode (get_ta (), Pmode); rtx ma = gen_int_mode (get_ma (), Pmode); diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index ea1d59d9cf2..a80bf8d1a74 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -1406,6 +1406,9 @@ riscv_v_adjust_bytesize (machine_mode mode, int scale) { if (riscv_v_ext_vector_mode_p (mode)) { + if (TARGET_XTHEADVECTOR) + return BYTES_PER_RISCV_VECTOR; + poly_int64 nunits = GET_MODE_NUNITS (mode); poly_int64 mode_size = GET_MODE_SIZE (mode); @@ -9965,7 +9968,7 @@ riscv_use_divmod_expander (void) static machine_mode riscv_preferred_simd_mode (scalar_mode mode) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::preferred_simd_mode (mode); return word_mode; @@ -10316,7 +10319,7 @@ riscv_mode_priority (int, int n) unsigned int riscv_autovectorize_vector_modes (vector_modes *modes, bool all) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::autovectorize_vector_modes (modes, all); return default_autovectorize_vector_modes (modes, all); @@ -10499,6 +10502,16 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) return false; } +/* Implements target hook vector_mode_supported_any_target_p. */ + +static bool +riscv_vector_mode_supported_any_target_p (machine_mode mode) +{ + if (TARGET_XTHEADVECTOR) + return false; + return true; +} + /* Initialize the GCC target structure. */ #undef TARGET_ASM_ALIGNED_HI_OP #define TARGET_ASM_ALIGNED_HI_OP "\t.half\t" @@ -10842,6 +10855,9 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) #undef TARGET_PREFERRED_ELSE_VALUE #define TARGET_PREFERRED_ELSE_VALUE riscv_preferred_else_value +#undef TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P +#define TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P riscv_vector_mode_supported_any_target_p + struct gcc_target targetm = TARGET_INITIALIZER; #include "gt-riscv.h" diff --git a/gcc/config/riscv/riscv_th_vector.h b/gcc/config/riscv/riscv_th_vector.h new file mode 100644 index 00000000000..6f47e0c90a4 --- /dev/null +++ b/gcc/config/riscv/riscv_th_vector.h @@ -0,0 +1,49 @@ +/* RISC-V 'XTheadVector' Extension intrinsics include file. + Copyright (C) 2022-2023 Free Software Foundation, Inc. + + This file is part of GCC. + + GCC is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published + by the Free Software Foundation; either version 3, or (at your + option) any later version. + + GCC is distributed in the hope that it will be useful, but WITHOUT + ANY WARRANTY; without even the implied warranty of MERCHANTABILITY + or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public + License for more details. + + Under Section 7 of GPL version 3, you are granted additional + permissions described in the GCC Runtime Library Exception, version + 3.1, as published by the Free Software Foundation. + + You should have received a copy of the GNU General Public License and + a copy of the GCC Runtime Library Exception along with this program; + see the files COPYING3 and COPYING.RUNTIME respectively. If not, see + <http://www.gnu.org/licenses/>. */ + +#ifndef __RISCV_TH_VECTOR_H +#define __RISCV_TH_VECTOR_H + +#include <stdint.h> +#include <stddef.h> + +#ifndef __riscv_xtheadvector +#error "XTheadVector intrinsics require the xtheadvector extension." +#else +#ifdef __cplusplus +extern "C" { +#endif + +/* NOTE: This implementation of riscv_th_vector.h is intentionally short. It does + not define the RVV types and intrinsic functions directly in C and C++ + code, but instead uses the following pragma to tell GCC to insert the + necessary type and function definitions itself. The net effect is the + same, and the file is a complete implementation of riscv_th_vector.h. */ +#pragma riscv intrinsic "vector" + +#ifdef __cplusplus +} +#endif // __cplusplus +#endif // __riscv_xtheadvector +#endif // __RISCV_TH_ECTOR_H diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md new file mode 100644 index 00000000000..f0aa196d93b --- /dev/null +++ b/gcc/config/riscv/thead-vector.md @@ -0,0 +1,142 @@ +(define_c_enum "unspec" [ + UNSPEC_TH_VWLDST +]) + +(define_mode_iterator V_VLS_VT [V VLS VT]) +(define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) + +(define_split + [(set (match_operand:V_VB_VLS_VT 0 "reg_or_mem_operand") + (match_operand:V_VB_VLS_VT 1 "reg_or_mem_operand"))] + "TARGET_XTHEADVECTOR" + [(const_int 0)] + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + }) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:V_VLS_VT 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:V_VLS_VT + [(match_operand:V_VLS_VT 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2")]) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:VB 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:VB + [(match_operand:VB 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2") + (set (attr "sew") (const_int 8)) + (set (attr "vlmul") (symbol_ref "riscv_vector::LMUL_1"))]) + +(define_insn "@th_vsetvl<mode>" + [(set (match_operand:P 0 "register_operand" "=r") + (unspec:P [(match_operand:P 1 "vector_length_operand" "rK") + (match_operand 2 "const_int_operand" "i") + (match_operand 3 "const_int_operand" "i")] UNSPEC_VSETVL)) + (set (reg:SI VL_REGNUM) + (unspec:SI [(match_dup 1) + (match_dup 2) + (match_dup 3)] UNSPEC_VSETVL)) + (set (reg:SI VTYPE_REGNUM) + (unspec:SI [(match_dup 2) + (match_dup 3)] UNSPEC_VSETVL))] + "TARGET_XTHEADVECTOR" + "vsetvli\t%0,%1,e%2,%m3" + [(set_attr "type" "vsetvl") + (set_attr "mode" "<MODE>") + (set (attr "sew") (symbol_ref "INTVAL (operands[2])")) + (set (attr "vlmul") (symbol_ref "INTVAL (operands[3])"))]) + +;; vsetvl zero,zero,vtype instruction. +;; This pattern has no side effects and does not set X0 register. +(define_insn "th_vsetvl_vtype_change_only" + [(set (reg:SI VTYPE_REGNUM) + (unspec:SI + [(match_operand 0 "const_int_operand" "i") + (match_operand 1 "const_int_operand" "i")] UNSPEC_VSETVL))] + "TARGET_XTHEADVECTOR" + "vsetvli\tzero,zero,e%0,%m1" + [(set_attr "type" "vsetvl") + (set_attr "mode" "SI") + (set (attr "sew") (symbol_ref "INTVAL (operands[0])")) + (set (attr "vlmul") (symbol_ref "INTVAL (operands[1])"))]) + +;; vsetvl zero,rs1,vtype instruction. +;; The reason we need this pattern since we should avoid setting X0 register +;; in vsetvl instruction pattern. +(define_insn "@th_vsetvl_discard_result<mode>" + [(set (reg:SI VL_REGNUM) + (unspec:SI [(match_operand:P 0 "vector_length_operand" "rK") + (match_operand 1 "const_int_operand" "i") + (match_operand 2 "const_int_operand" "i")] UNSPEC_VSETVL)) + (set (reg:SI VTYPE_REGNUM) + (unspec:SI [(match_dup 1) + (match_dup 2)] UNSPEC_VSETVL))] + "TARGET_XTHEADVECTOR" + "vsetvli\tzero,%0,e%1,%m2" + [(set_attr "type" "vsetvl") + (set_attr "mode" "<MODE>") + (set (attr "sew") (symbol_ref "INTVAL (operands[1])")) + (set (attr "vlmul") (symbol_ref "INTVAL (operands[2])"))]) + +;; It's emit by vsetvl/vsetvlmax intrinsics with no side effects. +;; Since we have many optmization passes from "expand" to "reload_completed", +;; such pattern can allow us gain benefits of these optimizations. +(define_insn_and_split "@th_vsetvl<mode>_no_side_effects" + [(set (match_operand:P 0 "register_operand" "=r") + (unspec:P [(match_operand:P 1 "vector_length_operand" "rK") + (match_operand 2 "const_int_operand" "i") + (match_operand 3 "const_int_operand" "i")] UNSPEC_VSETVL))] + "TARGET_XTHEADVECTOR" + "#" + "&& epilogue_completed" + [(parallel + [(set (match_dup 0) + (unspec:P [(match_dup 1) (match_dup 2) (match_dup 3)] UNSPEC_VSETVL)) + (set (reg:SI VL_REGNUM) + (unspec:SI [(match_dup 1) (match_dup 2) (match_dup 3)] UNSPEC_VSETVL)) + (set (reg:SI VTYPE_REGNUM) + (unspec:SI [(match_dup 2) (match_dup 3)] UNSPEC_VSETVL))])] + "" + [(set_attr "type" "vsetvl") + (set_attr "mode" "SI")]) diff --git a/gcc/config/riscv/vector-iterators.md b/gcc/config/riscv/vector-iterators.md index 5f5f7b5b986..c0fc7a2441d 100644 --- a/gcc/config/riscv/vector-iterators.md +++ b/gcc/config/riscv/vector-iterators.md @@ -109,11 +109,11 @@ ]) (define_mode_iterator VI [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -128,11 +128,11 @@ ;; allow the instruction and mode to be matched during combine et al. (define_mode_iterator VF [ (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -140,11 +140,11 @@ (define_mode_iterator VF_ZVFHMIN [ (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -271,16 +271,16 @@ ]) (define_mode_iterator VEEWEXT2 [ - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -290,10 +290,10 @@ ]) (define_mode_iterator VEEWEXT4 [ - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -311,59 +311,59 @@ ]) (define_mode_iterator VEEWTRUNC2 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") (RVVM4SI "TARGET_64BIT") (RVVM2SI "TARGET_64BIT") (RVVM1SI "TARGET_64BIT") - (RVVMF2SI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEEWTRUNC4 [ - RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM2HI "TARGET_64BIT") (RVVM1HI "TARGET_64BIT") - (RVVMF2HI "TARGET_64BIT") - (RVVMF4HI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2HI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") (RVVM1HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEEWTRUNC8 [ (RVVM1QI "TARGET_64BIT") - (RVVMF2QI "TARGET_64BIT") - (RVVMF4QI "TARGET_64BIT") - (RVVMF8QI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2QI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF4QI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEI16 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -452,11 +452,11 @@ ]) (define_mode_iterator VFULLI [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_FULL_V") (RVVM4DI "TARGET_FULL_V") (RVVM2DI "TARGET_FULL_V") (RVVM1DI "TARGET_FULL_V") @@ -509,17 +509,17 @@ ]) (define_mode_iterator VI_QH [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") ]) (define_mode_iterator VI_QHS [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -560,11 +560,11 @@ ]) (define_mode_iterator VI_QHS_NO_M8 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -603,11 +603,11 @@ (define_mode_iterator VF_HS [ (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (V1HF "riscv_vector::vls_mode_valid_p (V1HFmode) && TARGET_ZVFH") (V2HF "riscv_vector::vls_mode_valid_p (V2HFmode) && TARGET_ZVFH") @@ -638,12 +638,12 @@ (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") - (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (V1HF "riscv_vector::vls_mode_valid_p (V1HFmode) && TARGET_ZVFH") (V2HF "riscv_vector::vls_mode_valid_p (V2HFmode) && TARGET_ZVFH") @@ -674,11 +674,11 @@ ]) (define_mode_iterator V_VLSI_QHS [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -756,27 +756,27 @@ ;; E.g. when index mode = RVVM8QImde and Pmode = SImode, if it is not zero_extend or ;; scalar != 1, such gather/scatter is not allowed since we don't have RVVM32SImode. (define_mode_iterator RATIO64 [ - (RVVMF8QI "TARGET_MIN_VLEN > 32") - (RVVMF4HI "TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM1DI "TARGET_VECTOR_ELEN_64") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") ]) (define_mode_iterator RATIO32 [ - RVVMF4QI - RVVMF2HI + (RVVMF4QI "!TARGET_XTHEADVECTOR") + (RVVMF2HI "!TARGET_XTHEADVECTOR") RVVM1SI (RVVM2DI "TARGET_VECTOR_ELEN_64") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") ]) (define_mode_iterator RATIO16 [ - RVVMF2QI + (RVVMF2QI "!TARGET_XTHEADVECTOR") RVVM1HI RVVM2SI (RVVM4DI "TARGET_VECTOR_ELEN_64") @@ -814,21 +814,21 @@ ]) (define_mode_iterator RATIO64I [ - (RVVMF8QI "TARGET_MIN_VLEN > 32") - (RVVMF4HI "TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM1DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") ]) (define_mode_iterator RATIO32I [ - RVVMF4QI - RVVMF2HI + (RVVMF4QI "!TARGET_XTHEADVECTOR") + (RVVMF2HI "!TARGET_XTHEADVECTOR") RVVM1SI (RVVM2DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") ]) (define_mode_iterator RATIO16I [ - RVVMF2QI + (RVVMF2QI "!TARGET_XTHEADVECTOR") RVVM1HI RVVM2SI (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") @@ -873,21 +873,21 @@ ]) (define_mode_iterator V_FRACT [ - RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") ]) (define_mode_iterator VWEXTI [ - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -933,7 +933,7 @@ (RVVM4SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -966,7 +966,7 @@ (RVVM4SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -996,7 +996,7 @@ (define_mode_iterator VWCONVERTI [ (RVVM8SI "TARGET_ZVFH") (RVVM4SI "TARGET_ZVFH") (RVVM2SI "TARGET_ZVFH") (RVVM1SI "TARGET_ZVFH") - (RVVMF2SI "TARGET_ZVFH") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_ZVFH") (RVVM8DI "TARGET_VECTOR_ELEN_64 && TARGET_VECTOR_ELEN_FP_32") (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_VECTOR_ELEN_FP_32") @@ -1045,7 +1045,7 @@ ]) (define_mode_iterator VQEXTI [ - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -1456,11 +1456,11 @@ ;; VINDEXED [VI8 VI16 VI32 (VI64 "TARGET_64BIT")]. (define_mode_iterator VINDEXED [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") @@ -1468,12 +1468,12 @@ (RVVM1DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64 && TARGET_64BIT") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64 && TARGET_64BIT") @@ -3173,11 +3173,11 @@ (define_mode_iterator V_VLS_F_CONVERT_SI [ (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") - (RVVMF2HF "TARGET_ZVFH") (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") @@ -3290,12 +3290,12 @@ ]) (define_mode_iterator V_VLS_F_CONVERT_DI [ - (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index b5a9055cdc4..9e7a487759a 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -83,7 +83,7 @@ ;; check. However, we need default value of SEW for vsetvl instruction since there ;; is no field for ratio in the vsetvl instruction encoding. (define_attr "sew" "" - (cond [(eq_attr "mode" "RVVMF64BI,RVVMF32BI,RVVMF16BI,RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ + (cond [(eq_attr "mode" "RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ RVVM8QI,RVVM4QI,RVVM2QI,RVVM1QI,RVVMF2QI,RVVMF4QI,RVVMF8QI,\ RVVM1x8QI,RVVMF2x8QI,RVVMF4x8QI,RVVMF8x8QI,\ RVVM1x7QI,RVVMF2x7QI,RVVMF4x7QI,RVVMF8x7QI,\ @@ -95,6 +95,18 @@ V1QI,V2QI,V4QI,V8QI,V16QI,V32QI,V64QI,V128QI,V256QI,V512QI,V1024QI,V2048QI,V4096QI,\ V1BI,V2BI,V4BI,V8BI,V16BI,V32BI,V64BI,V128BI,V256BI,V512BI,V1024BI,V2048BI,V4096BI") (const_int 8) + (eq_attr "mode" "RVVMF16BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 16) + (const_int 8)) + (eq_attr "mode" "RVVMF32BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 32) + (const_int 8)) + (eq_attr "mode" "RVVMF64BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 64) + (const_int 8)) (eq_attr "mode" "RVVM8HI,RVVM4HI,RVVM2HI,RVVM1HI,RVVMF2HI,RVVMF4HI,\ RVVM1x8HI,RVVMF2x8HI,RVVMF4x8HI,\ RVVM1x7HI,RVVMF2x7HI,RVVMF4x7HI,\ @@ -155,9 +167,9 @@ (eq_attr "mode" "RVVM4QI,RVVMF2BI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2QI,RVVMF4BI") (symbol_ref "riscv_vector::LMUL_2") (eq_attr "mode" "RVVM1QI,RVVMF8BI") (symbol_ref "riscv_vector::LMUL_1") - (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "riscv_vector::LMUL_F2") - (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "riscv_vector::LMUL_F4") - (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "riscv_vector::LMUL_F8") + (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F2") + (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F4") + (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F8") (eq_attr "mode" "RVVM8HI") (symbol_ref "riscv_vector::LMUL_8") (eq_attr "mode" "RVVM4HI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2HI") (symbol_ref "riscv_vector::LMUL_2") @@ -428,6 +440,10 @@ vislide1up,vislide1down,vfslide1up,vfslide1down,\ vgather,vcompress,vlsegdux,vlsegdox,vssegtux,vssegtox") (const_int INVALID_ATTRIBUTE) + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE) (eq_attr "mode" "RVVM8QI,RVVM1BI") (const_int 1) (eq_attr "mode" "RVVM4QI,RVVMF2BI") (const_int 2) (eq_attr "mode" "RVVM2QI,RVVMF4BI") (const_int 4) @@ -888,6 +904,8 @@ (symbol_ref "riscv_vector::FRM_DYN")] (symbol_ref "riscv_vector::FRM_NONE"))) +(include "thead-vector.md") + ;; ----------------------------------------------------------------- ;; ---- Miscellaneous Operations ;; ----------------------------------------------------------------- @@ -1097,7 +1115,7 @@ (define_insn "*mov<mode>_whole" [(set (match_operand:V_WHOLE 0 "reg_or_mem_operand" "=vr, m,vr") (match_operand:V_WHOLE 1 "reg_or_mem_operand" " m,vr,vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "@ vl%m1re<sew>.v\t%0,%1 vs%m1r.v\t%1,%0 @@ -1125,7 +1143,7 @@ (define_insn "*mov<mode>" [(set (match_operand:VB 0 "register_operand" "=vr") (match_operand:VB 1 "register_operand" " vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "vmv1r.v\t%0,%1" [(set_attr "type" "vmov") (set_attr "mode" "<MODE>")]) @@ -3680,7 +3698,7 @@ (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3701,7 +3719,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3722,7 +3740,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c index 2e0e12aa045..2eef9e1e1a8 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c @@ -1,4 +1,4 @@ -/* { dg-do compile } */ +/* { dg-do compile { target { ! riscv_xtheadvector } } } */ /* { dg-skip-if "test rvv intrinsic" { *-*-* } { "*" } { "-march=rv*v*" } } */ void foo0 () {__rvv_bool64_t t;} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c index 3d81b179235..ef329e30785 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c @@ -1,4 +1,4 @@ /* { dg-do compile } */ /* { dg-options "-O3 -march=rv32gc -mabi=ilp32d" } */ -#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' extension enabled} } */ +#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' or 'XTHEADVECTOR' extension enabled} } */ diff --git a/gcc/testsuite/lib/target-supports.exp b/gcc/testsuite/lib/target-supports.exp index 7f13ff0ca56..70df6b1401c 100644 --- a/gcc/testsuite/lib/target-supports.exp +++ b/gcc/testsuite/lib/target-supports.exp @@ -1952,6 +1952,18 @@ proc check_effective_target_riscv_zbb { } { }] } +# Return 1 if the target arch supports the XTheadVector extension, 0 otherwise. +# Cache the result. + +proc check_effective_target_riscv_xtheadvector { } { + return [check_no_compiler_messages riscv_ext_xtheadvector assembly { + #ifndef __riscv_xtheadvector + #error "Not __riscv_xtheadvector" + #endif + }] +} + + # Return 1 if we can execute code when using dg-add-options riscv_v proc check_effective_target_riscv_v_ok { } { -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: [PATCH v4] RISC-V: Handle differences between XTheadvector and Vector 2023-12-29 4:21 ` [PATCH v4] RISC-V: Handle differences between XTheadvector and Vector Jun Sha (Joshua) @ 2024-01-02 2:00 ` juzhe.zhong 2024-01-02 3:03 ` Re：[PATCH " joshua 2024-01-02 9:48 ` joshua 2024-01-02 12:39 ` [PATCH " Jun Sha (Joshua) 1 sibling, 2 replies; 133+ messages in thread From: juzhe.zhong @ 2024-01-02 2:00 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, cooper.joshua, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 70404 bytes --] + if (TARGET_XTHEADVECTOR) + return false; Move it to : if (TARGET_VECTOR && stringop_strategy & STRATEGY_VECTOR) { bool ok = riscv_vector::expand_block_move (dest, src, length); if (ok) return true; } (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (match_test "!TARGET_XTHEADVECTOR || rtx_equal_p (op, const0_rtx)") + (match_operand 0 "const_csr_operand")))) It's hard to trace. Change it into : (ior 1. TARGET_THEADVECTOR && rtx_equal_p (op, const0_rtx) 2. !TAGEET_THEADVECTOR && const_csr_operand) + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (mode, dest, src, + RVV_VLMAX, GEN_INT(VLMAX))); + return true; + } Move it outside legitimize_move It should be it: if (TARGET_THEADVECTOR) emit_th_move... DONE; else if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; vsetvli issues: I wonder whether we can use ASM_OUTPUT_OPCODE to recognize "ta,ma"/"ta,mu"/"tu,ma"/"tu,mu" and replace these 4 variants by "". So that we don't have tail policy and mask policy in vsetvli ASM string. Another alternative approach is we can change vsetlvi ASM rule: "vset%i1vli\t%0,%1,e%2,%m3,t%p4,m%p5" if (TARGET_THEADVECTOR) ... else else if (code == CONST_INT) { /* Tail && Mask policy. */ asm_fprintf (file, "%s", IS_AGNOSTIC (UINTVAL (op)) ? "a" : "u"); } in riscv.cc. The benefit is that we can avoid adding all th_vsetvl patterns and invasive code changs in VSETVL PASS. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2023-12-29 12:21 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v4] RISC-V: Handle differences between XTheadvector and Vector This patch is to handle the differences in instruction generation between Vector and XTheadVector. In this version, we only support partial xtheadvector instructions that leverage directly from current RVV1.0 with simple adding "th." prefix. For different name xtheadvector instructions but share same patterns as RVV1.0 instructions, we will use ASM targethook to rewrite the whole string of the instructions in the following patches. For some vector patterns that cannot be avoided, we use "!TARGET_XTHEADVECTOR" to disable them in vector.md in order not to generate instructions that xtheadvector does not support, like vmv1r and vsext.vf2. gcc/ChangeLog: * config.gcc: Add files for XTheadVector intrinsics. * config/riscv/autovec.md: Guard XTheadVector. * config/riscv/riscv-string.cc (expand_block_move): Guard XTheadVector. * config/riscv/riscv-v.cc (legitimize_move): New expansion. (get_prefer_tail_policy): Give specific value for tail. (get_prefer_mask_policy): Give specific value for mask. (vls_mode_valid_p): Avoid autovec. * config/riscv/riscv-vector-builtins-shapes.cc (check_type): (build_one): New function. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): (DEF_THEAD_RVV_FUNCTION): Add new marcos. (check_required_extensions): (handle_pragma_vector): * config/riscv/riscv-vector-builtins.h (RVV_REQUIRE_VECTOR): (RVV_REQUIRE_XTHEADVECTOR): Add RVV_REQUIRE_VECTOR and RVV_REQUIRE_XTHEADVECTOR. (struct function_group_info): * config/riscv/riscv-vector-switch.def (ENTRY): Disable fractional mode for the XTheadVector extension. (TUPLE_ENTRY): Likewise. * config/riscv/riscv-vsetvl.cc: Add functions for xtheadvector. * config/riscv/riscv.cc (riscv_v_ext_vls_mode_p): Guard XTheadVector. (riscv_v_adjust_bytesize): Likewise. (riscv_preferred_simd_mode): Likewsie. (riscv_autovectorize_vector_modes): Likewise. (riscv_vector_mode_supported_any_target_p): Likewise. (TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P): Likewise. * config/riscv/vector-iterators.md: Remove fractional LMUL. * config/riscv/vector.md: Include thead-vector.md. * config/riscv/riscv_th_vector.h: New file. * config/riscv/thead-vector.md: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/base/pragma-1.c: Add XTheadVector. * gcc.target/riscv/rvv/base/abi-1.c: Exclude XTheadVector. * lib/target-supports.exp: Add target for XTheadVector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config.gcc | 2 +- gcc/config/riscv/autovec.md | 2 +- gcc/config/riscv/predicates.md | 3 +- gcc/config/riscv/riscv-string.cc | 3 + gcc/config/riscv/riscv-v.cc | 13 +- .../riscv/riscv-vector-builtins-bases.cc | 3 + .../riscv/riscv-vector-builtins-shapes.cc | 23 +++ gcc/config/riscv/riscv-vector-switch.def | 150 +++++++------- gcc/config/riscv/riscv-vsetvl.cc | 10 + gcc/config/riscv/riscv.cc | 20 +- gcc/config/riscv/riscv_th_vector.h | 49 +++++ gcc/config/riscv/thead-vector.md | 142 +++++++++++++ gcc/config/riscv/vector-iterators.md | 186 +++++++++--------- gcc/config/riscv/vector.md | 36 +++- .../gcc.target/riscv/rvv/base/abi-1.c | 2 +- .../gcc.target/riscv/rvv/base/pragma-1.c | 2 +- gcc/testsuite/lib/target-supports.exp | 12 ++ 17 files changed, 471 insertions(+), 187 deletions(-) create mode 100644 gcc/config/riscv/riscv_th_vector.h create mode 100644 gcc/config/riscv/thead-vector.md diff --git a/gcc/config.gcc b/gcc/config.gcc index f0676c830e8..1445d98c147 100644 --- a/gcc/config.gcc +++ b/gcc/config.gcc @@ -549,7 +549,7 @@ riscv*) extra_objs="${extra_objs} riscv-vector-builtins.o riscv-vector-builtins-shapes.o riscv-vector-builtins-bases.o" extra_objs="${extra_objs} thead.o riscv-target-attr.o" d_target_objs="riscv-d.o" - extra_headers="riscv_vector.h" + extra_headers="riscv_vector.h riscv_th_vector.h" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.cc" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.h" ;; diff --git a/gcc/config/riscv/autovec.md b/gcc/config/riscv/autovec.md index 8b8a92f10a1..1fac56c7095 100644 --- a/gcc/config/riscv/autovec.md +++ b/gcc/config/riscv/autovec.md @@ -2579,7 +2579,7 @@ [(match_operand 0 "register_operand") (match_operand 1 "memory_operand") (match_operand:ANYI 2 "const_int_operand")] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" { riscv_vector::expand_rawmemchr(<MODE>mode, operands[0], operands[1], operands[2]); diff --git a/gcc/config/riscv/predicates.md b/gcc/config/riscv/predicates.md index b87a6900841..c5ab86d5db9 100644 --- a/gcc/config/riscv/predicates.md +++ b/gcc/config/riscv/predicates.md @@ -428,7 +428,8 @@ ;; Predicates for the V extension. (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (match_test "!TARGET_XTHEADVECTOR || rtx_equal_p (op, const0_rtx)") + (match_operand 0 "const_csr_operand")))) (define_special_predicate "autovec_length_operand" (ior (match_operand 0 "pmode_register_operand") diff --git a/gcc/config/riscv/riscv-string.cc b/gcc/config/riscv/riscv-string.cc index 11c1f74d0b3..ec8f3486fd8 100644 --- a/gcc/config/riscv/riscv-string.cc +++ b/gcc/config/riscv/riscv-string.cc @@ -808,6 +808,9 @@ expand_block_move (rtx dst_in, rtx src_in, rtx length_in) bnez a2, loop # Any more? ret # Return */ + if (TARGET_XTHEADVECTOR) + return false; + gcc_assert (TARGET_VECTOR); HOST_WIDE_INT potential_ew diff --git a/gcc/config/riscv/riscv-v.cc b/gcc/config/riscv/riscv-v.cc index 038ab084a37..5e9e45aecd2 100644 --- a/gcc/config/riscv/riscv-v.cc +++ b/gcc/config/riscv/riscv-v.cc @@ -1523,6 +1523,13 @@ legitimize_move (rtx dest, rtx *srcp) return true; } + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (mode, dest, src, + RVV_VLMAX, GEN_INT(VLMAX))); + return true; + } + if (riscv_v_ext_vls_mode_p (mode)) { if (GET_MODE_NUNITS (mode).to_constant () <= 31) @@ -1772,7 +1779,7 @@ get_prefer_tail_policy () compiler pick up either agnostic or undisturbed. Maybe we will have a compile option like -mprefer=agnostic to set this value???. */ - return TAIL_ANY; + return TARGET_XTHEADVECTOR ? TAIL_AGNOSTIC : TAIL_ANY; } /* Get prefer mask policy. */ @@ -1783,7 +1790,7 @@ get_prefer_mask_policy () compiler pick up either agnostic or undisturbed. Maybe we will have a compile option like -mprefer=agnostic to set this value???. */ - return MASK_ANY; + return TARGET_XTHEADVECTOR ? MASK_UNDISTURBED : MASK_ANY; } /* Get avl_type rtx. */ @@ -4383,7 +4390,7 @@ cmp_lmul_gt_one (machine_mode mode) bool vls_mode_valid_p (machine_mode vls_mode) { - if (!TARGET_VECTOR) + if (!TARGET_VECTOR || TARGET_XTHEADVECTOR) return false; if (riscv_autovec_preference == RVV_SCALABLE) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index c51affde353..2918c07ebf3 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -133,6 +133,9 @@ public: = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + if (TARGET_XTHEADVECTOR) + return e.generate_insn (code_for_th_vsetvl_no_side_effects (Pmode)); + /* TAIL_ANY. */ e.add_input_operand (Pmode, gen_int_mode (get_prefer_tail_policy (), Pmode)); diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 4a754e0228f..6b49404a1fa 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -33,6 +33,25 @@ namespace riscv_vector { +/* Check whether the RETURN_TYPE and ARGUMENT_TYPES are + valid for the function. */ + +static bool +check_type (tree return_type, vec<tree> &argument_types) +{ + tree arg; + unsigned i; + + if (!return_type) + return false; + + FOR_EACH_VEC_ELT (argument_types, i, arg) + if (!arg) + return false; + + return true; +} + /* Add one function instance for GROUP, using operand suffix at index OI, mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ static void @@ -49,6 +68,10 @@ build_one (function_builder &b, const function_group_info &group, group.ops_infos.types[vec_type_idx].index); b.allocate_argument_types (function_instance, argument_types); b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + b.add_overloaded_function (function_instance, *group.shape); b.add_unique_function (function_instance, (*group.shape), return_type, argument_types); diff --git a/gcc/config/riscv/riscv-vector-switch.def b/gcc/config/riscv/riscv-vector-switch.def index 5c9f9bcbc3e..f7a66b34bae 100644 --- a/gcc/config/riscv/riscv-vector-switch.def +++ b/gcc/config/riscv/riscv-vector-switch.def @@ -68,9 +68,9 @@ Encode the ratio of SEW/LMUL into the mask types. #endif /* Disable modes if TARGET_MIN_VLEN == 32. */ -ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) -ENTRY (RVVMF32BI, true, LMUL_F4, 32) -ENTRY (RVVMF16BI, true, LMUL_F2, 16) +ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F8, 64) +ENTRY (RVVMF32BI, true, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F4, 32) +ENTRY (RVVMF16BI, true, TARGET_XTHEADVECTOR ? LMUL_1 : LMUL_F2 , 16) ENTRY (RVVMF8BI, true, LMUL_1, 8) ENTRY (RVVMF4BI, true, LMUL_2, 4) ENTRY (RVVMF2BI, true, LMUL_4, 2) @@ -81,39 +81,39 @@ ENTRY (RVVM8QI, true, LMUL_8, 1) ENTRY (RVVM4QI, true, LMUL_4, 2) ENTRY (RVVM2QI, true, LMUL_2, 4) ENTRY (RVVM1QI, true, LMUL_1, 8) -ENTRY (RVVMF2QI, true, LMUL_F2, 16) -ENTRY (RVVMF4QI, true, LMUL_F4, 32) -ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) +ENTRY (RVVMF2QI, !TARGET_XTHEADVECTOR, LMUL_F2, 16) +ENTRY (RVVMF4QI, !TARGET_XTHEADVECTOR, LMUL_F4, 32) +ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F8, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8HI, true, LMUL_8, 2) ENTRY (RVVM4HI, true, LMUL_4, 4) ENTRY (RVVM2HI, true, LMUL_2, 8) ENTRY (RVVM1HI, true, LMUL_1, 16) -ENTRY (RVVMF2HI, true, LMUL_F2, 32) -ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HI, !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_16. */ ENTRY (RVVM8HF, TARGET_VECTOR_ELEN_FP_16, LMUL_8, 2) ENTRY (RVVM4HF, TARGET_VECTOR_ELEN_FP_16, LMUL_4, 4) ENTRY (RVVM2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_2, 8) ENTRY (RVVM1HF, TARGET_VECTOR_ELEN_FP_16, LMUL_1, 16) -ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_F2, 32) -ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8SI, true, LMUL_8, 4) ENTRY (RVVM4SI, true, LMUL_4, 8) ENTRY (RVVM2SI, true, LMUL_2, 16) ENTRY (RVVM1SI, true, LMUL_1, 32) -ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_32. */ ENTRY (RVVM8SF, TARGET_VECTOR_ELEN_FP_32, LMUL_8, 4) ENTRY (RVVM4SF, TARGET_VECTOR_ELEN_FP_32, LMUL_4, 8) ENTRY (RVVM2SF, TARGET_VECTOR_ELEN_FP_32, LMUL_2, 16) ENTRY (RVVM1SF, TARGET_VECTOR_ELEN_FP_32, LMUL_1, 32) -ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if !TARGET_VECTOR_ELEN_64. */ ENTRY (RVVM8DI, TARGET_VECTOR_ELEN_64, LMUL_8, 8) @@ -140,127 +140,127 @@ ENTRY (RVVM1DF, TARGET_VECTOR_ELEN_FP_64, LMUL_1, 64) #endif TUPLE_ENTRY (RVVM1x8QI, true, RVVM1QI, 8, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x8QI, true, RVVMF2QI, 8, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x8QI, true, RVVMF4QI, 8, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 8, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x8QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 8, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x8QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 8, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 8, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x7QI, true, RVVM1QI, 7, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x7QI, true, RVVMF2QI, 7, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x7QI, true, RVVMF4QI, 7, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 7, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x7QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 7, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x7QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 7, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 7, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x6QI, true, RVVM1QI, 6, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x6QI, true, RVVMF2QI, 6, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x6QI, true, RVVMF4QI, 6, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 6, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x6QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 6, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x6QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 6, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 6, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x5QI, true, RVVM1QI, 5, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x5QI, true, RVVMF2QI, 5, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x5QI, true, RVVMF4QI, 5, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 5, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x5QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 5, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x5QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 5, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 5, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x4QI, true, RVVM2QI, 4, LMUL_2, 4) TUPLE_ENTRY (RVVM1x4QI, true, RVVM1QI, 4, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x4QI, true, RVVMF2QI, 4, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x4QI, true, RVVMF4QI, 4, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 4, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x4QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 4, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x4QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 4, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 4, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x3QI, true, RVVM2QI, 3, LMUL_2, 4) TUPLE_ENTRY (RVVM1x3QI, true, RVVM1QI, 3, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x3QI, true, RVVMF2QI, 3, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x3QI, true, RVVMF4QI, 3, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 3, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x3QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 3, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x3QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 3, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 3, LMUL_F8, 64) TUPLE_ENTRY (RVVM4x2QI, true, RVVM4QI, 2, LMUL_4, 2) TUPLE_ENTRY (RVVM2x2QI, true, RVVM2QI, 2, LMUL_2, 4) TUPLE_ENTRY (RVVM1x2QI, true, RVVM1QI, 2, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x2QI, true, RVVMF2QI, 2, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x2QI, true, RVVMF4QI, 2, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 2, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x2QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 2, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x2QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 2, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 2, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x8HI, true, RVVM1HI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HI, true, RVVMF2HI, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HI, true, RVVM1HI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HI, true, RVVMF2HI, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HI, true, RVVM1HI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HI, true, RVVMF2HI, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HI, true, RVVM1HI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HI, true, RVVMF2HI, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HI, true, RVVM2HI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HI, true, RVVM1HI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HI, true, RVVMF2HI, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HI, true, RVVM2HI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HI, true, RVVM1HI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HI, true, RVVMF2HI, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HI, true, RVVM4HI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HI, true, RVVM2HI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HI, true, RVVM1HI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HI, true, RVVMF2HI, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM4HF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8SI, true, RVVM1SI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SI, true, RVVM1SI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SI, true, RVVM1SI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SI, true, RVVM1SI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SI, true, RVVM2SI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SI, true, RVVM1SI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SI, true, RVVM2SI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SI, true, RVVM1SI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SI, true, RVVM4SI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SI, true, RVVM2SI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SI, true, RVVM1SI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM4SF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 8, LMUL_1, 16) TUPLE_ENTRY (RVVM1x7DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 7, LMUL_1, 16) diff --git a/gcc/config/riscv/riscv-vsetvl.cc b/gcc/config/riscv/riscv-vsetvl.cc index eabaef80f89..c726253c107 100644 --- a/gcc/config/riscv/riscv-vsetvl.cc +++ b/gcc/config/riscv/riscv-vsetvl.cc @@ -1117,6 +1117,16 @@ public: avl = GEN_INT (0); rtx sew = gen_int_mode (get_sew (), Pmode); rtx vlmul = gen_int_mode (get_vlmul (), Pmode); + + if (TARGET_XTHEADVECTOR) { + if (change_vtype_only_p ()) + return gen_th_vsetvl_vtype_change_only (sew, vlmul); + else if (has_vl () && !ignore_vl) + return gen_th_vsetvl (Pmode, get_vl (), avl, sew, vlmul); + else + return gen_th_vsetvl_discard_result (Pmode, avl, sew, vlmul); + } + rtx ta = gen_int_mode (get_ta (), Pmode); rtx ma = gen_int_mode (get_ma (), Pmode); diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index ea1d59d9cf2..a80bf8d1a74 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -1406,6 +1406,9 @@ riscv_v_adjust_bytesize (machine_mode mode, int scale) { if (riscv_v_ext_vector_mode_p (mode)) { + if (TARGET_XTHEADVECTOR) + return BYTES_PER_RISCV_VECTOR; + poly_int64 nunits = GET_MODE_NUNITS (mode); poly_int64 mode_size = GET_MODE_SIZE (mode); @@ -9965,7 +9968,7 @@ riscv_use_divmod_expander (void) static machine_mode riscv_preferred_simd_mode (scalar_mode mode) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::preferred_simd_mode (mode); return word_mode; @@ -10316,7 +10319,7 @@ riscv_mode_priority (int, int n) unsigned int riscv_autovectorize_vector_modes (vector_modes *modes, bool all) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::autovectorize_vector_modes (modes, all); return default_autovectorize_vector_modes (modes, all); @@ -10499,6 +10502,16 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) return false; } +/* Implements target hook vector_mode_supported_any_target_p. */ + +static bool +riscv_vector_mode_supported_any_target_p (machine_mode mode) +{ + if (TARGET_XTHEADVECTOR) + return false; + return true; +} + /* Initialize the GCC target structure. */ #undef TARGET_ASM_ALIGNED_HI_OP #define TARGET_ASM_ALIGNED_HI_OP "\t.half\t" @@ -10842,6 +10855,9 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) #undef TARGET_PREFERRED_ELSE_VALUE #define TARGET_PREFERRED_ELSE_VALUE riscv_preferred_else_value +#undef TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P +#define TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P riscv_vector_mode_supported_any_target_p + struct gcc_target targetm = TARGET_INITIALIZER; #include "gt-riscv.h" diff --git a/gcc/config/riscv/riscv_th_vector.h b/gcc/config/riscv/riscv_th_vector.h new file mode 100644 index 00000000000..6f47e0c90a4 --- /dev/null +++ b/gcc/config/riscv/riscv_th_vector.h @@ -0,0 +1,49 @@ +/* RISC-V 'XTheadVector' Extension intrinsics include file. + Copyright (C) 2022-2023 Free Software Foundation, Inc. + + This file is part of GCC. + + GCC is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published + by the Free Software Foundation; either version 3, or (at your + option) any later version. + + GCC is distributed in the hope that it will be useful, but WITHOUT + ANY WARRANTY; without even the implied warranty of MERCHANTABILITY + or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public + License for more details. + + Under Section 7 of GPL version 3, you are granted additional + permissions described in the GCC Runtime Library Exception, version + 3.1, as published by the Free Software Foundation. + + You should have received a copy of the GNU General Public License and + a copy of the GCC Runtime Library Exception along with this program; + see the files COPYING3 and COPYING.RUNTIME respectively. If not, see + <http://www.gnu.org/licenses/>. */ + +#ifndef __RISCV_TH_VECTOR_H +#define __RISCV_TH_VECTOR_H + +#include <stdint.h> +#include <stddef.h> + +#ifndef __riscv_xtheadvector +#error "XTheadVector intrinsics require the xtheadvector extension." +#else +#ifdef __cplusplus +extern "C" { +#endif + +/* NOTE: This implementation of riscv_th_vector.h is intentionally short. It does + not define the RVV types and intrinsic functions directly in C and C++ + code, but instead uses the following pragma to tell GCC to insert the + necessary type and function definitions itself. The net effect is the + same, and the file is a complete implementation of riscv_th_vector.h. */ +#pragma riscv intrinsic "vector" + +#ifdef __cplusplus +} +#endif // __cplusplus +#endif // __riscv_xtheadvector +#endif // __RISCV_TH_ECTOR_H diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md new file mode 100644 index 00000000000..f0aa196d93b --- /dev/null +++ b/gcc/config/riscv/thead-vector.md @@ -0,0 +1,142 @@ +(define_c_enum "unspec" [ + UNSPEC_TH_VWLDST +]) + +(define_mode_iterator V_VLS_VT [V VLS VT]) +(define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) + +(define_split + [(set (match_operand:V_VB_VLS_VT 0 "reg_or_mem_operand") + (match_operand:V_VB_VLS_VT 1 "reg_or_mem_operand"))] + "TARGET_XTHEADVECTOR" + [(const_int 0)] + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + }) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:V_VLS_VT 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:V_VLS_VT + [(match_operand:V_VLS_VT 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2")]) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:VB 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:VB + [(match_operand:VB 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2") + (set (attr "sew") (const_int 8)) + (set (attr "vlmul") (symbol_ref "riscv_vector::LMUL_1"))]) + +(define_insn "@th_vsetvl<mode>" + [(set (match_operand:P 0 "register_operand" "=r") + (unspec:P [(match_operand:P 1 "vector_length_operand" "rK") + (match_operand 2 "const_int_operand" "i") + (match_operand 3 "const_int_operand" "i")] UNSPEC_VSETVL)) + (set (reg:SI VL_REGNUM) + (unspec:SI [(match_dup 1) + (match_dup 2) + (match_dup 3)] UNSPEC_VSETVL)) + (set (reg:SI VTYPE_REGNUM) + (unspec:SI [(match_dup 2) + (match_dup 3)] UNSPEC_VSETVL))] + "TARGET_XTHEADVECTOR" + "vsetvli\t%0,%1,e%2,%m3" + [(set_attr "type" "vsetvl") + (set_attr "mode" "<MODE>") + (set (attr "sew") (symbol_ref "INTVAL (operands[2])")) + (set (attr "vlmul") (symbol_ref "INTVAL (operands[3])"))]) + +;; vsetvl zero,zero,vtype instruction. +;; This pattern has no side effects and does not set X0 register. +(define_insn "th_vsetvl_vtype_change_only" + [(set (reg:SI VTYPE_REGNUM) + (unspec:SI + [(match_operand 0 "const_int_operand" "i") + (match_operand 1 "const_int_operand" "i")] UNSPEC_VSETVL))] + "TARGET_XTHEADVECTOR" + "vsetvli\tzero,zero,e%0,%m1" + [(set_attr "type" "vsetvl") + (set_attr "mode" "SI") + (set (attr "sew") (symbol_ref "INTVAL (operands[0])")) + (set (attr "vlmul") (symbol_ref "INTVAL (operands[1])"))]) + +;; vsetvl zero,rs1,vtype instruction. +;; The reason we need this pattern since we should avoid setting X0 register +;; in vsetvl instruction pattern. +(define_insn "@th_vsetvl_discard_result<mode>" + [(set (reg:SI VL_REGNUM) + (unspec:SI [(match_operand:P 0 "vector_length_operand" "rK") + (match_operand 1 "const_int_operand" "i") + (match_operand 2 "const_int_operand" "i")] UNSPEC_VSETVL)) + (set (reg:SI VTYPE_REGNUM) + (unspec:SI [(match_dup 1) + (match_dup 2)] UNSPEC_VSETVL))] + "TARGET_XTHEADVECTOR" + "vsetvli\tzero,%0,e%1,%m2" + [(set_attr "type" "vsetvl") + (set_attr "mode" "<MODE>") + (set (attr "sew") (symbol_ref "INTVAL (operands[1])")) + (set (attr "vlmul") (symbol_ref "INTVAL (operands[2])"))]) + +;; It's emit by vsetvl/vsetvlmax intrinsics with no side effects. +;; Since we have many optmization passes from "expand" to "reload_completed", +;; such pattern can allow us gain benefits of these optimizations. +(define_insn_and_split "@th_vsetvl<mode>_no_side_effects" + [(set (match_operand:P 0 "register_operand" "=r") + (unspec:P [(match_operand:P 1 "vector_length_operand" "rK") + (match_operand 2 "const_int_operand" "i") + (match_operand 3 "const_int_operand" "i")] UNSPEC_VSETVL))] + "TARGET_XTHEADVECTOR" + "#" + "&& epilogue_completed" + [(parallel + [(set (match_dup 0) + (unspec:P [(match_dup 1) (match_dup 2) (match_dup 3)] UNSPEC_VSETVL)) + (set (reg:SI VL_REGNUM) + (unspec:SI [(match_dup 1) (match_dup 2) (match_dup 3)] UNSPEC_VSETVL)) + (set (reg:SI VTYPE_REGNUM) + (unspec:SI [(match_dup 2) (match_dup 3)] UNSPEC_VSETVL))])] + "" + [(set_attr "type" "vsetvl") + (set_attr "mode" "SI")]) diff --git a/gcc/config/riscv/vector-iterators.md b/gcc/config/riscv/vector-iterators.md index 5f5f7b5b986..c0fc7a2441d 100644 --- a/gcc/config/riscv/vector-iterators.md +++ b/gcc/config/riscv/vector-iterators.md @@ -109,11 +109,11 @@ ]) (define_mode_iterator VI [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -128,11 +128,11 @@ ;; allow the instruction and mode to be matched during combine et al. (define_mode_iterator VF [ (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -140,11 +140,11 @@ (define_mode_iterator VF_ZVFHMIN [ (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -271,16 +271,16 @@ ]) (define_mode_iterator VEEWEXT2 [ - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -290,10 +290,10 @@ ]) (define_mode_iterator VEEWEXT4 [ - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -311,59 +311,59 @@ ]) (define_mode_iterator VEEWTRUNC2 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") (RVVM4SI "TARGET_64BIT") (RVVM2SI "TARGET_64BIT") (RVVM1SI "TARGET_64BIT") - (RVVMF2SI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEEWTRUNC4 [ - RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM2HI "TARGET_64BIT") (RVVM1HI "TARGET_64BIT") - (RVVMF2HI "TARGET_64BIT") - (RVVMF4HI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2HI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") (RVVM1HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEEWTRUNC8 [ (RVVM1QI "TARGET_64BIT") - (RVVMF2QI "TARGET_64BIT") - (RVVMF4QI "TARGET_64BIT") - (RVVMF8QI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2QI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF4QI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEI16 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -452,11 +452,11 @@ ]) (define_mode_iterator VFULLI [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_FULL_V") (RVVM4DI "TARGET_FULL_V") (RVVM2DI "TARGET_FULL_V") (RVVM1DI "TARGET_FULL_V") @@ -509,17 +509,17 @@ ]) (define_mode_iterator VI_QH [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") ]) (define_mode_iterator VI_QHS [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -560,11 +560,11 @@ ]) (define_mode_iterator VI_QHS_NO_M8 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -603,11 +603,11 @@ (define_mode_iterator VF_HS [ (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (V1HF "riscv_vector::vls_mode_valid_p (V1HFmode) && TARGET_ZVFH") (V2HF "riscv_vector::vls_mode_valid_p (V2HFmode) && TARGET_ZVFH") @@ -638,12 +638,12 @@ (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") - (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (V1HF "riscv_vector::vls_mode_valid_p (V1HFmode) && TARGET_ZVFH") (V2HF "riscv_vector::vls_mode_valid_p (V2HFmode) && TARGET_ZVFH") @@ -674,11 +674,11 @@ ]) (define_mode_iterator V_VLSI_QHS [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -756,27 +756,27 @@ ;; E.g. when index mode = RVVM8QImde and Pmode = SImode, if it is not zero_extend or ;; scalar != 1, such gather/scatter is not allowed since we don't have RVVM32SImode. (define_mode_iterator RATIO64 [ - (RVVMF8QI "TARGET_MIN_VLEN > 32") - (RVVMF4HI "TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM1DI "TARGET_VECTOR_ELEN_64") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") ]) (define_mode_iterator RATIO32 [ - RVVMF4QI - RVVMF2HI + (RVVMF4QI "!TARGET_XTHEADVECTOR") + (RVVMF2HI "!TARGET_XTHEADVECTOR") RVVM1SI (RVVM2DI "TARGET_VECTOR_ELEN_64") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") ]) (define_mode_iterator RATIO16 [ - RVVMF2QI + (RVVMF2QI "!TARGET_XTHEADVECTOR") RVVM1HI RVVM2SI (RVVM4DI "TARGET_VECTOR_ELEN_64") @@ -814,21 +814,21 @@ ]) (define_mode_iterator RATIO64I [ - (RVVMF8QI "TARGET_MIN_VLEN > 32") - (RVVMF4HI "TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM1DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") ]) (define_mode_iterator RATIO32I [ - RVVMF4QI - RVVMF2HI + (RVVMF4QI "!TARGET_XTHEADVECTOR") + (RVVMF2HI "!TARGET_XTHEADVECTOR") RVVM1SI (RVVM2DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") ]) (define_mode_iterator RATIO16I [ - RVVMF2QI + (RVVMF2QI "!TARGET_XTHEADVECTOR") RVVM1HI RVVM2SI (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") @@ -873,21 +873,21 @@ ]) (define_mode_iterator V_FRACT [ - RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") ]) (define_mode_iterator VWEXTI [ - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -933,7 +933,7 @@ (RVVM4SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -966,7 +966,7 @@ (RVVM4SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -996,7 +996,7 @@ (define_mode_iterator VWCONVERTI [ (RVVM8SI "TARGET_ZVFH") (RVVM4SI "TARGET_ZVFH") (RVVM2SI "TARGET_ZVFH") (RVVM1SI "TARGET_ZVFH") - (RVVMF2SI "TARGET_ZVFH") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_ZVFH") (RVVM8DI "TARGET_VECTOR_ELEN_64 && TARGET_VECTOR_ELEN_FP_32") (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_VECTOR_ELEN_FP_32") @@ -1045,7 +1045,7 @@ ]) (define_mode_iterator VQEXTI [ - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -1456,11 +1456,11 @@ ;; VINDEXED [VI8 VI16 VI32 (VI64 "TARGET_64BIT")]. (define_mode_iterator VINDEXED [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") @@ -1468,12 +1468,12 @@ (RVVM1DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64 && TARGET_64BIT") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64 && TARGET_64BIT") @@ -3173,11 +3173,11 @@ (define_mode_iterator V_VLS_F_CONVERT_SI [ (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") - (RVVMF2HF "TARGET_ZVFH") (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") @@ -3290,12 +3290,12 @@ ]) (define_mode_iterator V_VLS_F_CONVERT_DI [ - (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index b5a9055cdc4..9e7a487759a 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -83,7 +83,7 @@ ;; check. However, we need default value of SEW for vsetvl instruction since there ;; is no field for ratio in the vsetvl instruction encoding. (define_attr "sew" "" - (cond [(eq_attr "mode" "RVVMF64BI,RVVMF32BI,RVVMF16BI,RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ + (cond [(eq_attr "mode" "RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ RVVM8QI,RVVM4QI,RVVM2QI,RVVM1QI,RVVMF2QI,RVVMF4QI,RVVMF8QI,\ RVVM1x8QI,RVVMF2x8QI,RVVMF4x8QI,RVVMF8x8QI,\ RVVM1x7QI,RVVMF2x7QI,RVVMF4x7QI,RVVMF8x7QI,\ @@ -95,6 +95,18 @@ V1QI,V2QI,V4QI,V8QI,V16QI,V32QI,V64QI,V128QI,V256QI,V512QI,V1024QI,V2048QI,V4096QI,\ V1BI,V2BI,V4BI,V8BI,V16BI,V32BI,V64BI,V128BI,V256BI,V512BI,V1024BI,V2048BI,V4096BI") (const_int 8) + (eq_attr "mode" "RVVMF16BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 16) + (const_int 8)) + (eq_attr "mode" "RVVMF32BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 32) + (const_int 8)) + (eq_attr "mode" "RVVMF64BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 64) + (const_int 8)) (eq_attr "mode" "RVVM8HI,RVVM4HI,RVVM2HI,RVVM1HI,RVVMF2HI,RVVMF4HI,\ RVVM1x8HI,RVVMF2x8HI,RVVMF4x8HI,\ RVVM1x7HI,RVVMF2x7HI,RVVMF4x7HI,\ @@ -155,9 +167,9 @@ (eq_attr "mode" "RVVM4QI,RVVMF2BI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2QI,RVVMF4BI") (symbol_ref "riscv_vector::LMUL_2") (eq_attr "mode" "RVVM1QI,RVVMF8BI") (symbol_ref "riscv_vector::LMUL_1") - (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "riscv_vector::LMUL_F2") - (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "riscv_vector::LMUL_F4") - (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "riscv_vector::LMUL_F8") + (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F2") + (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F4") + (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F8") (eq_attr "mode" "RVVM8HI") (symbol_ref "riscv_vector::LMUL_8") (eq_attr "mode" "RVVM4HI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2HI") (symbol_ref "riscv_vector::LMUL_2") @@ -428,6 +440,10 @@ vislide1up,vislide1down,vfslide1up,vfslide1down,\ vgather,vcompress,vlsegdux,vlsegdox,vssegtux,vssegtox") (const_int INVALID_ATTRIBUTE) + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE) (eq_attr "mode" "RVVM8QI,RVVM1BI") (const_int 1) (eq_attr "mode" "RVVM4QI,RVVMF2BI") (const_int 2) (eq_attr "mode" "RVVM2QI,RVVMF4BI") (const_int 4) @@ -888,6 +904,8 @@ (symbol_ref "riscv_vector::FRM_DYN")] (symbol_ref "riscv_vector::FRM_NONE"))) +(include "thead-vector.md") + ;; ----------------------------------------------------------------- ;; ---- Miscellaneous Operations ;; ----------------------------------------------------------------- @@ -1097,7 +1115,7 @@ (define_insn "*mov<mode>_whole" [(set (match_operand:V_WHOLE 0 "reg_or_mem_operand" "=vr, m,vr") (match_operand:V_WHOLE 1 "reg_or_mem_operand" " m,vr,vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "@ vl%m1re<sew>.v\t%0,%1 vs%m1r.v\t%1,%0 @@ -1125,7 +1143,7 @@ (define_insn "*mov<mode>" [(set (match_operand:VB 0 "register_operand" "=vr") (match_operand:VB 1 "register_operand" " vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "vmv1r.v\t%0,%1" [(set_attr "type" "vmov") (set_attr "mode" "<MODE>")]) @@ -3680,7 +3698,7 @@ (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3701,7 +3719,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3722,7 +3740,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c index 2e0e12aa045..2eef9e1e1a8 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c @@ -1,4 +1,4 @@ -/* { dg-do compile } */ +/* { dg-do compile { target { ! riscv_xtheadvector } } } */ /* { dg-skip-if "test rvv intrinsic" { *-*-* } { "*" } { "-march=rv*v*" } } */ void foo0 () {__rvv_bool64_t t;} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c index 3d81b179235..ef329e30785 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c @@ -1,4 +1,4 @@ /* { dg-do compile } */ /* { dg-options "-O3 -march=rv32gc -mabi=ilp32d" } */ -#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' extension enabled} } */ +#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' or 'XTHEADVECTOR' extension enabled} } */ diff --git a/gcc/testsuite/lib/target-supports.exp b/gcc/testsuite/lib/target-supports.exp index 7f13ff0ca56..70df6b1401c 100644 --- a/gcc/testsuite/lib/target-supports.exp +++ b/gcc/testsuite/lib/target-supports.exp @@ -1952,6 +1952,18 @@ proc check_effective_target_riscv_zbb { } { }] } +# Return 1 if the target arch supports the XTheadVector extension, 0 otherwise. +# Cache the result. + +proc check_effective_target_riscv_xtheadvector { } { + return [check_no_compiler_messages riscv_ext_xtheadvector assembly { + #ifndef __riscv_xtheadvector + #error "Not __riscv_xtheadvector" + #endif + }] +} + + # Return 1 if we can execute code when using dg-add-options riscv_v proc check_effective_target_riscv_v_ok { } { -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re：[PATCH v4] RISC-V: Handle differences between XTheadvector and Vector 2024-01-02 2:00 ` juzhe.zhong @ 2024-01-02 3:03 ` joshua 2024-01-02 3:10 ` juzhe.zhong 2024-01-02 9:48 ` joshua 1 sibling, 1 reply; 133+ messages in thread From: joshua @ 2024-01-02 3:03 UTC (permalink / raw) To: juzhe.zhong, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner For vsetvl issues, what we want to do here is to directly remove "t" and "m". If we add TARGET_XTHEADVECTOR logic in case "p" in riscv_print_operand, how can we remove "t" and "m"? If I use "break", assembly like "th.vsetvli zero,a5,e8,m1,t,m" will be returned. if (TARGET_THEADVECTOR) ... else else if (code == CONST_INT) { /* Tail && Mask policy. */ asm_fprintf (file, "%s", IS_AGNOSTIC (UINTVAL (op)) ? "a" : "u"); } ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月2日(星期二) 10:00 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v4] RISC-V: Handle differences between XTheadvector and Vector + if (TARGET_XTHEADVECTOR) + return false; Move it to : if (TARGET_VECTOR && stringop_strategy & STRATEGY_VECTOR) { bool ok = riscv_vector::expand_block_move (dest, src, length); if (ok) return true; } (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (match_test "!TARGET_XTHEADVECTOR || rtx_equal_p (op, const0_rtx)") + (match_operand 0 "const_csr_operand")))) It's hard to trace. Change it into : (ior 1. TARGET_THEADVECTOR && rtx_equal_p (op, const0_rtx) 2. !TAGEET_THEADVECTOR && const_csr_operand) + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (mode, dest, src, + RVV_VLMAX, GEN_INT(VLMAX))); + return true; + } Move it outside legitimize_move It should be it: if (TARGET_THEADVECTOR) emit_th_move... DONE; else if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; vsetvli issues: I wonder whether we can use ASM_OUTPUT_OPCODE to recognize "ta,ma"/"ta,mu"/"tu,ma"/"tu,mu" and replace these 4 variants by "". So that we don't have tail policy and mask policy in vsetvli ASM string. Another alternative approach is we can change vsetlvi ASM rule: "vset%i1vli\t%0,%1,e%2,%m3,t%p4,m%p5" if (TARGET_THEADVECTOR) ... else else if (code == CONST_INT) { /* Tail && Mask policy. */ asm_fprintf (file, "%s", IS_AGNOSTIC (UINTVAL (op)) ? "a" : "u"); } in riscv.cc. The benefit is that we can avoid adding all th_vsetvl patterns and invasive code changs in VSETVL PASS. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2023-12-29 12:21 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v4] RISC-V: Handle differences between XTheadvector and Vector This patch is to handle the differences in instruction generation between Vector and XTheadVector. In this version, we only support partial xtheadvector instructions that leverage directly from current RVV1.0 with simple adding "th." prefix. For different name xtheadvector instructions but share same patterns as RVV1.0 instructions, we will use ASM targethook to rewrite the whole string of the instructions in the following patches. For some vector patterns that cannot be avoided, we use "!TARGET_XTHEADVECTOR" to disable them in vector.md in order not to generate instructions that xtheadvector does not support, like vmv1r and vsext.vf2. gcc/ChangeLog: * config.gcc: Add files for XTheadVector intrinsics. * config/riscv/autovec.md: Guard XTheadVector. * config/riscv/riscv-string.cc (expand_block_move): Guard XTheadVector. * config/riscv/riscv-v.cc (legitimize_move): New expansion. (get_prefer_tail_policy): Give specific value for tail. (get_prefer_mask_policy): Give specific value for mask. (vls_mode_valid_p): Avoid autovec. * config/riscv/riscv-vector-builtins-shapes.cc (check_type): (build_one): New function. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): (DEF_THEAD_RVV_FUNCTION): Add new marcos. (check_required_extensions): (handle_pragma_vector): * config/riscv/riscv-vector-builtins.h (RVV_REQUIRE_VECTOR): (RVV_REQUIRE_XTHEADVECTOR): Add RVV_REQUIRE_VECTOR and RVV_REQUIRE_XTHEADVECTOR. (struct function_group_info): * config/riscv/riscv-vector-switch.def (ENTRY): Disable fractional mode for the XTheadVector extension. (TUPLE_ENTRY): Likewise. * config/riscv/riscv-vsetvl.cc: Add functions for xtheadvector. * config/riscv/riscv.cc (riscv_v_ext_vls_mode_p): Guard XTheadVector. (riscv_v_adjust_bytesize): Likewise. (riscv_preferred_simd_mode): Likewsie. (riscv_autovectorize_vector_modes): Likewise. (riscv_vector_mode_supported_any_target_p): Likewise. (TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P): Likewise. * config/riscv/vector-iterators.md: Remove fractional LMUL. * config/riscv/vector.md: Include thead-vector.md. * config/riscv/riscv_th_vector.h: New file. * config/riscv/thead-vector.md: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/base/pragma-1.c: Add XTheadVector. * gcc.target/riscv/rvv/base/abi-1.c: Exclude XTheadVector. * lib/target-supports.exp: Add target for XTheadVector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config.gcc | 2 +- gcc/config/riscv/autovec.md | 2 +- gcc/config/riscv/predicates.md | 3 +- gcc/config/riscv/riscv-string.cc | 3 + gcc/config/riscv/riscv-v.cc | 13 +- .../riscv/riscv-vector-builtins-bases.cc | 3 + .../riscv/riscv-vector-builtins-shapes.cc | 23 +++ gcc/config/riscv/riscv-vector-switch.def | 150 +++++++------- gcc/config/riscv/riscv-vsetvl.cc | 10 + gcc/config/riscv/riscv.cc | 20 +- gcc/config/riscv/riscv_th_vector.h | 49 +++++ gcc/config/riscv/thead-vector.md | 142 +++++++++++++ gcc/config/riscv/vector-iterators.md | 186 +++++++++--------- gcc/config/riscv/vector.md | 36 +++- .../gcc.target/riscv/rvv/base/abi-1.c | 2 +- .../gcc.target/riscv/rvv/base/pragma-1.c | 2 +- gcc/testsuite/lib/target-supports.exp | 12 ++ 17 files changed, 471 insertions(+), 187 deletions(-) create mode 100644 gcc/config/riscv/riscv_th_vector.h create mode 100644 gcc/config/riscv/thead-vector.md diff --git a/gcc/config.gcc b/gcc/config.gcc index f0676c830e8..1445d98c147 100644 --- a/gcc/config.gcc +++ b/gcc/config.gcc @@ -549,7 +549,7 @@ riscv*) extra_objs="${extra_objs} riscv-vector-builtins.o riscv-vector-builtins-shapes.o riscv-vector-builtins-bases.o" extra_objs="${extra_objs} thead.o riscv-target-attr.o" d_target_objs="riscv-d.o" - extra_headers="riscv_vector.h" + extra_headers="riscv_vector.h riscv_th_vector.h" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.cc" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.h" ;; diff --git a/gcc/config/riscv/autovec.md b/gcc/config/riscv/autovec.md index 8b8a92f10a1..1fac56c7095 100644 --- a/gcc/config/riscv/autovec.md +++ b/gcc/config/riscv/autovec.md @@ -2579,7 +2579,7 @@ [(match_operand 0 "register_operand") (match_operand 1 "memory_operand") (match_operand:ANYI 2 "const_int_operand")] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" { riscv_vector::expand_rawmemchr(<MODE>mode, operands[0], operands[1], operands[2]); diff --git a/gcc/config/riscv/predicates.md b/gcc/config/riscv/predicates.md index b87a6900841..c5ab86d5db9 100644 --- a/gcc/config/riscv/predicates.md +++ b/gcc/config/riscv/predicates.md @@ -428,7 +428,8 @@ ;; Predicates for the V extension. (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (match_test "!TARGET_XTHEADVECTOR || rtx_equal_p (op, const0_rtx)") + (match_operand 0 "const_csr_operand")))) (define_special_predicate "autovec_length_operand" (ior (match_operand 0 "pmode_register_operand") diff --git a/gcc/config/riscv/riscv-string.cc b/gcc/config/riscv/riscv-string.cc index 11c1f74d0b3..ec8f3486fd8 100644 --- a/gcc/config/riscv/riscv-string.cc +++ b/gcc/config/riscv/riscv-string.cc @@ -808,6 +808,9 @@ expand_block_move (rtx dst_in, rtx src_in, rtx length_in) bnez a2, loop # Any more? ret # Return */ + if (TARGET_XTHEADVECTOR) + return false; + gcc_assert (TARGET_VECTOR); HOST_WIDE_INT potential_ew diff --git a/gcc/config/riscv/riscv-v.cc b/gcc/config/riscv/riscv-v.cc index 038ab084a37..5e9e45aecd2 100644 --- a/gcc/config/riscv/riscv-v.cc +++ b/gcc/config/riscv/riscv-v.cc @@ -1523,6 +1523,13 @@ legitimize_move (rtx dest, rtx *srcp) return true; } + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (mode, dest, src, + RVV_VLMAX, GEN_INT(VLMAX))); + return true; + } + if (riscv_v_ext_vls_mode_p (mode)) { if (GET_MODE_NUNITS (mode).to_constant () <= 31) @@ -1772,7 +1779,7 @@ get_prefer_tail_policy () compiler pick up either agnostic or undisturbed. Maybe we will have a compile option like -mprefer=agnostic to set this value???. */ - return TAIL_ANY; + return TARGET_XTHEADVECTOR ? TAIL_AGNOSTIC : TAIL_ANY; } /* Get prefer mask policy. */ @@ -1783,7 +1790,7 @@ get_prefer_mask_policy () compiler pick up either agnostic or undisturbed. Maybe we will have a compile option like -mprefer=agnostic to set this value???. */ - return MASK_ANY; + return TARGET_XTHEADVECTOR ? MASK_UNDISTURBED : MASK_ANY; } /* Get avl_type rtx. */ @@ -4383,7 +4390,7 @@ cmp_lmul_gt_one (machine_mode mode) bool vls_mode_valid_p (machine_mode vls_mode) { - if (!TARGET_VECTOR) + if (!TARGET_VECTOR || TARGET_XTHEADVECTOR) return false; if (riscv_autovec_preference == RVV_SCALABLE) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index c51affde353..2918c07ebf3 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -133,6 +133,9 @@ public: = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + if (TARGET_XTHEADVECTOR) + return e.generate_insn (code_for_th_vsetvl_no_side_effects (Pmode)); + /* TAIL_ANY. */ e.add_input_operand (Pmode, gen_int_mode (get_prefer_tail_policy (), Pmode)); diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 4a754e0228f..6b49404a1fa 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -33,6 +33,25 @@ namespace riscv_vector { +/* Check whether the RETURN_TYPE and ARGUMENT_TYPES are + valid for the function. */ + +static bool +check_type (tree return_type, vec<tree> &argument_types) +{ + tree arg; + unsigned i; + + if (!return_type) + return false; + + FOR_EACH_VEC_ELT (argument_types, i, arg) + if (!arg) + return false; + + return true; +} + /* Add one function instance for GROUP, using operand suffix at index OI, mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ static void @@ -49,6 +68,10 @@ build_one (function_builder &b, const function_group_info &group, group.ops_infos.types[vec_type_idx].index); b.allocate_argument_types (function_instance, argument_types); b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + b.add_overloaded_function (function_instance, *group.shape); b.add_unique_function (function_instance, (*group.shape), return_type, argument_types); diff --git a/gcc/config/riscv/riscv-vector-switch.def b/gcc/config/riscv/riscv-vector-switch.def index 5c9f9bcbc3e..f7a66b34bae 100644 --- a/gcc/config/riscv/riscv-vector-switch.def +++ b/gcc/config/riscv/riscv-vector-switch.def @@ -68,9 +68,9 @@ Encode the ratio of SEW/LMUL into the mask types. #endif /* Disable modes if TARGET_MIN_VLEN == 32. */ -ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) -ENTRY (RVVMF32BI, true, LMUL_F4, 32) -ENTRY (RVVMF16BI, true, LMUL_F2, 16) +ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F8, 64) +ENTRY (RVVMF32BI, true, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F4, 32) +ENTRY (RVVMF16BI, true, TARGET_XTHEADVECTOR ? LMUL_1 : LMUL_F2 , 16) ENTRY (RVVMF8BI, true, LMUL_1, 8) ENTRY (RVVMF4BI, true, LMUL_2, 4) ENTRY (RVVMF2BI, true, LMUL_4, 2) @@ -81,39 +81,39 @@ ENTRY (RVVM8QI, true, LMUL_8, 1) ENTRY (RVVM4QI, true, LMUL_4, 2) ENTRY (RVVM2QI, true, LMUL_2, 4) ENTRY (RVVM1QI, true, LMUL_1, 8) -ENTRY (RVVMF2QI, true, LMUL_F2, 16) -ENTRY (RVVMF4QI, true, LMUL_F4, 32) -ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) +ENTRY (RVVMF2QI, !TARGET_XTHEADVECTOR, LMUL_F2, 16) +ENTRY (RVVMF4QI, !TARGET_XTHEADVECTOR, LMUL_F4, 32) +ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F8, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8HI, true, LMUL_8, 2) ENTRY (RVVM4HI, true, LMUL_4, 4) ENTRY (RVVM2HI, true, LMUL_2, 8) ENTRY (RVVM1HI, true, LMUL_1, 16) -ENTRY (RVVMF2HI, true, LMUL_F2, 32) -ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HI, !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_16. */ ENTRY (RVVM8HF, TARGET_VECTOR_ELEN_FP_16, LMUL_8, 2) ENTRY (RVVM4HF, TARGET_VECTOR_ELEN_FP_16, LMUL_4, 4) ENTRY (RVVM2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_2, 8) ENTRY (RVVM1HF, TARGET_VECTOR_ELEN_FP_16, LMUL_1, 16) -ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_F2, 32) -ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8SI, true, LMUL_8, 4) ENTRY (RVVM4SI, true, LMUL_4, 8) ENTRY (RVVM2SI, true, LMUL_2, 16) ENTRY (RVVM1SI, true, LMUL_1, 32) -ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_32. */ ENTRY (RVVM8SF, TARGET_VECTOR_ELEN_FP_32, LMUL_8, 4) ENTRY (RVVM4SF, TARGET_VECTOR_ELEN_FP_32, LMUL_4, 8) ENTRY (RVVM2SF, TARGET_VECTOR_ELEN_FP_32, LMUL_2, 16) ENTRY (RVVM1SF, TARGET_VECTOR_ELEN_FP_32, LMUL_1, 32) -ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if !TARGET_VECTOR_ELEN_64. */ ENTRY (RVVM8DI, TARGET_VECTOR_ELEN_64, LMUL_8, 8) @@ -140,127 +140,127 @@ ENTRY (RVVM1DF, TARGET_VECTOR_ELEN_FP_64, LMUL_1, 64) #endif TUPLE_ENTRY (RVVM1x8QI, true, RVVM1QI, 8, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x8QI, true, RVVMF2QI, 8, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x8QI, true, RVVMF4QI, 8, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 8, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x8QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 8, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x8QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 8, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 8, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x7QI, true, RVVM1QI, 7, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x7QI, true, RVVMF2QI, 7, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x7QI, true, RVVMF4QI, 7, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 7, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x7QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 7, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x7QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 7, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 7, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x6QI, true, RVVM1QI, 6, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x6QI, true, RVVMF2QI, 6, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x6QI, true, RVVMF4QI, 6, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 6, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x6QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 6, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x6QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 6, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 6, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x5QI, true, RVVM1QI, 5, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x5QI, true, RVVMF2QI, 5, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x5QI, true, RVVMF4QI, 5, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 5, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x5QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 5, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x5QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 5, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 5, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x4QI, true, RVVM2QI, 4, LMUL_2, 4) TUPLE_ENTRY (RVVM1x4QI, true, RVVM1QI, 4, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x4QI, true, RVVMF2QI, 4, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x4QI, true, RVVMF4QI, 4, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 4, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x4QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 4, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x4QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 4, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 4, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x3QI, true, RVVM2QI, 3, LMUL_2, 4) TUPLE_ENTRY (RVVM1x3QI, true, RVVM1QI, 3, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x3QI, true, RVVMF2QI, 3, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x3QI, true, RVVMF4QI, 3, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 3, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x3QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 3, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x3QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 3, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 3, LMUL_F8, 64) TUPLE_ENTRY (RVVM4x2QI, true, RVVM4QI, 2, LMUL_4, 2) TUPLE_ENTRY (RVVM2x2QI, true, RVVM2QI, 2, LMUL_2, 4) TUPLE_ENTRY (RVVM1x2QI, true, RVVM1QI, 2, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x2QI, true, RVVMF2QI, 2, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x2QI, true, RVVMF4QI, 2, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 2, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x2QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 2, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x2QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 2, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 2, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x8HI, true, RVVM1HI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HI, true, RVVMF2HI, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HI, true, RVVM1HI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HI, true, RVVMF2HI, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HI, true, RVVM1HI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HI, true, RVVMF2HI, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HI, true, RVVM1HI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HI, true, RVVMF2HI, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HI, true, RVVM2HI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HI, true, RVVM1HI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HI, true, RVVMF2HI, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HI, true, RVVM2HI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HI, true, RVVM1HI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HI, true, RVVMF2HI, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HI, true, RVVM4HI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HI, true, RVVM2HI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HI, true, RVVM1HI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HI, true, RVVMF2HI, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM4HF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8SI, true, RVVM1SI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SI, true, RVVM1SI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SI, true, RVVM1SI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SI, true, RVVM1SI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SI, true, RVVM2SI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SI, true, RVVM1SI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SI, true, RVVM2SI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SI, true, RVVM1SI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SI, true, RVVM4SI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SI, true, RVVM2SI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SI, true, RVVM1SI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM4SF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 8, LMUL_1, 16) TUPLE_ENTRY (RVVM1x7DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 7, LMUL_1, 16) diff --git a/gcc/config/riscv/riscv-vsetvl.cc b/gcc/config/riscv/riscv-vsetvl.cc index eabaef80f89..c726253c107 100644 --- a/gcc/config/riscv/riscv-vsetvl.cc +++ b/gcc/config/riscv/riscv-vsetvl.cc @@ -1117,6 +1117,16 @@ public: avl = GEN_INT (0); rtx sew = gen_int_mode (get_sew (), Pmode); rtx vlmul = gen_int_mode (get_vlmul (), Pmode); + + if (TARGET_XTHEADVECTOR) { + if (change_vtype_only_p ()) + return gen_th_vsetvl_vtype_change_only (sew, vlmul); + else if (has_vl () && !ignore_vl) + return gen_th_vsetvl (Pmode, get_vl (), avl, sew, vlmul); + else + return gen_th_vsetvl_discard_result (Pmode, avl, sew, vlmul); + } + rtx ta = gen_int_mode (get_ta (), Pmode); rtx ma = gen_int_mode (get_ma (), Pmode); diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index ea1d59d9cf2..a80bf8d1a74 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -1406,6 +1406,9 @@ riscv_v_adjust_bytesize (machine_mode mode, int scale) { if (riscv_v_ext_vector_mode_p (mode)) { + if (TARGET_XTHEADVECTOR) + return BYTES_PER_RISCV_VECTOR; + poly_int64 nunits = GET_MODE_NUNITS (mode); poly_int64 mode_size = GET_MODE_SIZE (mode); @@ -9965,7 +9968,7 @@ riscv_use_divmod_expander (void) static machine_mode riscv_preferred_simd_mode (scalar_mode mode) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::preferred_simd_mode (mode); return word_mode; @@ -10316,7 +10319,7 @@ riscv_mode_priority (int, int n) unsigned int riscv_autovectorize_vector_modes (vector_modes *modes, bool all) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::autovectorize_vector_modes (modes, all); return default_autovectorize_vector_modes (modes, all); @@ -10499,6 +10502,16 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) return false; } +/* Implements target hook vector_mode_supported_any_target_p. */ + +static bool +riscv_vector_mode_supported_any_target_p (machine_mode mode) +{ + if (TARGET_XTHEADVECTOR) + return false; + return true; +} + /* Initialize the GCC target structure. */ #undef TARGET_ASM_ALIGNED_HI_OP #define TARGET_ASM_ALIGNED_HI_OP "\t.half\t" @@ -10842,6 +10855,9 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) #undef TARGET_PREFERRED_ELSE_VALUE #define TARGET_PREFERRED_ELSE_VALUE riscv_preferred_else_value +#undef TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P +#define TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P riscv_vector_mode_supported_any_target_p + struct gcc_target targetm = TARGET_INITIALIZER; #include "gt-riscv.h" diff --git a/gcc/config/riscv/riscv_th_vector.h b/gcc/config/riscv/riscv_th_vector.h new file mode 100644 index 00000000000..6f47e0c90a4 --- /dev/null +++ b/gcc/config/riscv/riscv_th_vector.h @@ -0,0 +1,49 @@ +/* RISC-V 'XTheadVector' Extension intrinsics include file. + Copyright (C) 2022-2023 Free Software Foundation, Inc. + + This file is part of GCC. + + GCC is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published + by the Free Software Foundation; either version 3, or (at your + option) any later version. + + GCC is distributed in the hope that it will be useful, but WITHOUT + ANY WARRANTY; without even the implied warranty of MERCHANTABILITY + or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public + License for more details. + + Under Section 7 of GPL version 3, you are granted additional + permissions described in the GCC Runtime Library Exception, version + 3.1, as published by the Free Software Foundation. + + You should have received a copy of the GNU General Public License and + a copy of the GCC Runtime Library Exception along with this program; + see the files COPYING3 and COPYING.RUNTIME respectively. If not, see + <" target="_blank">http://www.gnu.org/licenses/>. */ + +#ifndef __RISCV_TH_VECTOR_H +#define __RISCV_TH_VECTOR_H + +#include <stdint.h> +#include <stddef.h> + +#ifndef __riscv_xtheadvector +#error "XTheadVector intrinsics require the xtheadvector extension." +#else +#ifdef __cplusplus +extern "C" { +#endif + +/* NOTE: This implementation of riscv_th_vector.h is intentionally short. It does + not define the RVV types and intrinsic functions directly in C and C++ + code, but instead uses the following pragma to tell GCC to insert the + necessary type and function definitions itself. The net effect is the + same, and the file is a complete implementation of riscv_th_vector.h. */ +#pragma riscv intrinsic "vector" + +#ifdef __cplusplus +} +#endif // __cplusplus +#endif // __riscv_xtheadvector +#endif // __RISCV_TH_ECTOR_H diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md new file mode 100644 index 00000000000..f0aa196d93b --- /dev/null +++ b/gcc/config/riscv/thead-vector.md @@ -0,0 +1,142 @@ +(define_c_enum "unspec" [ + UNSPEC_TH_VWLDST +]) + +(define_mode_iterator V_VLS_VT [V VLS VT]) +(define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) + +(define_split + [(set (match_operand:V_VB_VLS_VT 0 "reg_or_mem_operand") + (match_operand:V_VB_VLS_VT 1 "reg_or_mem_operand"))] + "TARGET_XTHEADVECTOR" + [(const_int 0)] + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + }) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:V_VLS_VT 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:V_VLS_VT + [(match_operand:V_VLS_VT 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2")]) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:VB 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:VB + [(match_operand:VB 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2") + (set (attr "sew") (const_int 8)) + (set (attr "vlmul") (symbol_ref "riscv_vector::LMUL_1"))]) + +(define_insn "@th_vsetvl<mode>" + [(set (match_operand:P 0 "register_operand" "=r") + (unspec:P [(match_operand:P 1 "vector_length_operand" "rK") + (match_operand 2 "const_int_operand" "i") + (match_operand 3 "const_int_operand" "i")] UNSPEC_VSETVL)) + (set (reg:SI VL_REGNUM) + (unspec:SI [(match_dup 1) + (match_dup 2) + (match_dup 3)] UNSPEC_VSETVL)) + (set (reg:SI VTYPE_REGNUM) + (unspec:SI [(match_dup 2) + (match_dup 3)] UNSPEC_VSETVL))] + "TARGET_XTHEADVECTOR" + "vsetvli\t%0,%1,e%2,%m3" + [(set_attr "type" "vsetvl") + (set_attr "mode" "<MODE>") + (set (attr "sew") (symbol_ref "INTVAL (operands[2])")) + (set (attr "vlmul") (symbol_ref "INTVAL (operands[3])"))]) + +;; vsetvl zero,zero,vtype instruction. +;; This pattern has no side effects and does not set X0 register. +(define_insn "th_vsetvl_vtype_change_only" + [(set (reg:SI VTYPE_REGNUM) + (unspec:SI + [(match_operand 0 "const_int_operand" "i") + (match_operand 1 "const_int_operand" "i")] UNSPEC_VSETVL))] + "TARGET_XTHEADVECTOR" + "vsetvli\tzero,zero,e%0,%m1" + [(set_attr "type" "vsetvl") + (set_attr "mode" "SI") + (set (attr "sew") (symbol_ref "INTVAL (operands[0])")) + (set (attr "vlmul") (symbol_ref "INTVAL (operands[1])"))]) + +;; vsetvl zero,rs1,vtype instruction. +;; The reason we need this pattern since we should avoid setting X0 register +;; in vsetvl instruction pattern. +(define_insn "@th_vsetvl_discard_result<mode>" + [(set (reg:SI VL_REGNUM) + (unspec:SI [(match_operand:P 0 "vector_length_operand" "rK") + (match_operand 1 "const_int_operand" "i") + (match_operand 2 "const_int_operand" "i")] UNSPEC_VSETVL)) + (set (reg:SI VTYPE_REGNUM) + (unspec:SI [(match_dup 1) + (match_dup 2)] UNSPEC_VSETVL))] + "TARGET_XTHEADVECTOR" + "vsetvli\tzero,%0,e%1,%m2" + [(set_attr "type" "vsetvl") + (set_attr "mode" "<MODE>") + (set (attr "sew") (symbol_ref "INTVAL (operands[1])")) + (set (attr "vlmul") (symbol_ref "INTVAL (operands[2])"))]) + +;; It's emit by vsetvl/vsetvlmax intrinsics with no side effects. +;; Since we have many optmization passes from "expand" to "reload_completed", +;; such pattern can allow us gain benefits of these optimizations. +(define_insn_and_split "@th_vsetvl<mode>_no_side_effects" + [(set (match_operand:P 0 "register_operand" "=r") + (unspec:P [(match_operand:P 1 "vector_length_operand" "rK") + (match_operand 2 "const_int_operand" "i") + (match_operand 3 "const_int_operand" "i")] UNSPEC_VSETVL))] + "TARGET_XTHEADVECTOR" + "#" + "&& epilogue_completed" + [(parallel + [(set (match_dup 0) + (unspec:P [(match_dup 1) (match_dup 2) (match_dup 3)] UNSPEC_VSETVL)) + (set (reg:SI VL_REGNUM) + (unspec:SI [(match_dup 1) (match_dup 2) (match_dup 3)] UNSPEC_VSETVL)) + (set (reg:SI VTYPE_REGNUM) + (unspec:SI [(match_dup 2) (match_dup 3)] UNSPEC_VSETVL))])] + "" + [(set_attr "type" "vsetvl") + (set_attr "mode" "SI")]) diff --git a/gcc/config/riscv/vector-iterators.md b/gcc/config/riscv/vector-iterators.md index 5f5f7b5b986..c0fc7a2441d 100644 --- a/gcc/config/riscv/vector-iterators.md +++ b/gcc/config/riscv/vector-iterators.md @@ -109,11 +109,11 @@ ]) (define_mode_iterator VI [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -128,11 +128,11 @@ ;; allow the instruction and mode to be matched during combine et al. (define_mode_iterator VF [ (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -140,11 +140,11 @@ (define_mode_iterator VF_ZVFHMIN [ (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -271,16 +271,16 @@ ]) (define_mode_iterator VEEWEXT2 [ - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -290,10 +290,10 @@ ]) (define_mode_iterator VEEWEXT4 [ - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -311,59 +311,59 @@ ]) (define_mode_iterator VEEWTRUNC2 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") (RVVM4SI "TARGET_64BIT") (RVVM2SI "TARGET_64BIT") (RVVM1SI "TARGET_64BIT") - (RVVMF2SI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEEWTRUNC4 [ - RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM2HI "TARGET_64BIT") (RVVM1HI "TARGET_64BIT") - (RVVMF2HI "TARGET_64BIT") - (RVVMF4HI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2HI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") (RVVM1HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEEWTRUNC8 [ (RVVM1QI "TARGET_64BIT") - (RVVMF2QI "TARGET_64BIT") - (RVVMF4QI "TARGET_64BIT") - (RVVMF8QI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2QI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF4QI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEI16 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -452,11 +452,11 @@ ]) (define_mode_iterator VFULLI [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_FULL_V") (RVVM4DI "TARGET_FULL_V") (RVVM2DI "TARGET_FULL_V") (RVVM1DI "TARGET_FULL_V") @@ -509,17 +509,17 @@ ]) (define_mode_iterator VI_QH [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") ]) (define_mode_iterator VI_QHS [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -560,11 +560,11 @@ ]) (define_mode_iterator VI_QHS_NO_M8 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -603,11 +603,11 @@ (define_mode_iterator VF_HS [ (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (V1HF "riscv_vector::vls_mode_valid_p (V1HFmode) && TARGET_ZVFH") (V2HF "riscv_vector::vls_mode_valid_p (V2HFmode) && TARGET_ZVFH") @@ -638,12 +638,12 @@ (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") - (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (V1HF "riscv_vector::vls_mode_valid_p (V1HFmode) && TARGET_ZVFH") (V2HF "riscv_vector::vls_mode_valid_p (V2HFmode) && TARGET_ZVFH") @@ -674,11 +674,11 @@ ]) (define_mode_iterator V_VLSI_QHS [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -756,27 +756,27 @@ ;; E.g. when index mode = RVVM8QImde and Pmode = SImode, if it is not zero_extend or ;; scalar != 1, such gather/scatter is not allowed since we don't have RVVM32SImode. (define_mode_iterator RATIO64 [ - (RVVMF8QI "TARGET_MIN_VLEN > 32") - (RVVMF4HI "TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM1DI "TARGET_VECTOR_ELEN_64") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") ]) (define_mode_iterator RATIO32 [ - RVVMF4QI - RVVMF2HI + (RVVMF4QI "!TARGET_XTHEADVECTOR") + (RVVMF2HI "!TARGET_XTHEADVECTOR") RVVM1SI (RVVM2DI "TARGET_VECTOR_ELEN_64") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") ]) (define_mode_iterator RATIO16 [ - RVVMF2QI + (RVVMF2QI "!TARGET_XTHEADVECTOR") RVVM1HI RVVM2SI (RVVM4DI "TARGET_VECTOR_ELEN_64") @@ -814,21 +814,21 @@ ]) (define_mode_iterator RATIO64I [ - (RVVMF8QI "TARGET_MIN_VLEN > 32") - (RVVMF4HI "TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM1DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") ]) (define_mode_iterator RATIO32I [ - RVVMF4QI - RVVMF2HI + (RVVMF4QI "!TARGET_XTHEADVECTOR") + (RVVMF2HI "!TARGET_XTHEADVECTOR") RVVM1SI (RVVM2DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") ]) (define_mode_iterator RATIO16I [ - RVVMF2QI + (RVVMF2QI "!TARGET_XTHEADVECTOR") RVVM1HI RVVM2SI (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") @@ -873,21 +873,21 @@ ]) (define_mode_iterator V_FRACT [ - RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") ]) (define_mode_iterator VWEXTI [ - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -933,7 +933,7 @@ (RVVM4SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -966,7 +966,7 @@ (RVVM4SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -996,7 +996,7 @@ (define_mode_iterator VWCONVERTI [ (RVVM8SI "TARGET_ZVFH") (RVVM4SI "TARGET_ZVFH") (RVVM2SI "TARGET_ZVFH") (RVVM1SI "TARGET_ZVFH") - (RVVMF2SI "TARGET_ZVFH") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_ZVFH") (RVVM8DI "TARGET_VECTOR_ELEN_64 && TARGET_VECTOR_ELEN_FP_32") (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_VECTOR_ELEN_FP_32") @@ -1045,7 +1045,7 @@ ]) (define_mode_iterator VQEXTI [ - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -1456,11 +1456,11 @@ ;; VINDEXED [VI8 VI16 VI32 (VI64 "TARGET_64BIT")]. (define_mode_iterator VINDEXED [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") @@ -1468,12 +1468,12 @@ (RVVM1DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64 && TARGET_64BIT") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64 && TARGET_64BIT") @@ -3173,11 +3173,11 @@ (define_mode_iterator V_VLS_F_CONVERT_SI [ (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") - (RVVMF2HF "TARGET_ZVFH") (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") @@ -3290,12 +3290,12 @@ ]) (define_mode_iterator V_VLS_F_CONVERT_DI [ - (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index b5a9055cdc4..9e7a487759a 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -83,7 +83,7 @@ ;; check. However, we need default value of SEW for vsetvl instruction since there ;; is no field for ratio in the vsetvl instruction encoding. (define_attr "sew" "" - (cond [(eq_attr "mode" "RVVMF64BI,RVVMF32BI,RVVMF16BI,RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ + (cond [(eq_attr "mode" "RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ RVVM8QI,RVVM4QI,RVVM2QI,RVVM1QI,RVVMF2QI,RVVMF4QI,RVVMF8QI,\ RVVM1x8QI,RVVMF2x8QI,RVVMF4x8QI,RVVMF8x8QI,\ RVVM1x7QI,RVVMF2x7QI,RVVMF4x7QI,RVVMF8x7QI,\ @@ -95,6 +95,18 @@ V1QI,V2QI,V4QI,V8QI,V16QI,V32QI,V64QI,V128QI,V256QI,V512QI,V1024QI,V2048QI,V4096QI,\ V1BI,V2BI,V4BI,V8BI,V16BI,V32BI,V64BI,V128BI,V256BI,V512BI,V1024BI,V2048BI,V4096BI") (const_int 8) + (eq_attr "mode" "RVVMF16BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 16) + (const_int 8)) + (eq_attr "mode" "RVVMF32BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 32) + (const_int 8)) + (eq_attr "mode" "RVVMF64BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 64) + (const_int 8)) (eq_attr "mode" "RVVM8HI,RVVM4HI,RVVM2HI,RVVM1HI,RVVMF2HI,RVVMF4HI,\ RVVM1x8HI,RVVMF2x8HI,RVVMF4x8HI,\ RVVM1x7HI,RVVMF2x7HI,RVVMF4x7HI,\ @@ -155,9 +167,9 @@ (eq_attr "mode" "RVVM4QI,RVVMF2BI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2QI,RVVMF4BI") (symbol_ref "riscv_vector::LMUL_2") (eq_attr "mode" "RVVM1QI,RVVMF8BI") (symbol_ref "riscv_vector::LMUL_1") - (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "riscv_vector::LMUL_F2") - (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "riscv_vector::LMUL_F4") - (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "riscv_vector::LMUL_F8") + (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F2") + (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F4") + (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F8") (eq_attr "mode" "RVVM8HI") (symbol_ref "riscv_vector::LMUL_8") (eq_attr "mode" "RVVM4HI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2HI") (symbol_ref "riscv_vector::LMUL_2") @@ -428,6 +440,10 @@ vislide1up,vislide1down,vfslide1up,vfslide1down,\ vgather,vcompress,vlsegdux,vlsegdox,vssegtux,vssegtox") (const_int INVALID_ATTRIBUTE) + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE) (eq_attr "mode" "RVVM8QI,RVVM1BI") (const_int 1) (eq_attr "mode" "RVVM4QI,RVVMF2BI") (const_int 2) (eq_attr "mode" "RVVM2QI,RVVMF4BI") (const_int 4) @@ -888,6 +904,8 @@ (symbol_ref "riscv_vector::FRM_DYN")] (symbol_ref "riscv_vector::FRM_NONE"))) +(include "thead-vector.md") + ;; ----------------------------------------------------------------- ;; ---- Miscellaneous Operations ;; ----------------------------------------------------------------- @@ -1097,7 +1115,7 @@ (define_insn "*mov<mode>_whole" [(set (match_operand:V_WHOLE 0 "reg_or_mem_operand" "=vr, m,vr") (match_operand:V_WHOLE 1 "reg_or_mem_operand" " m,vr,vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "@ vl%m1re<sew>.v\t%0,%1 vs%m1r.v\t%1,%0 @@ -1125,7 +1143,7 @@ (define_insn "*mov<mode>" [(set (match_operand:VB 0 "register_operand" "=vr") (match_operand:VB 1 "register_operand" " vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "vmv1r.v\t%0,%1" [(set_attr "type" "vmov") (set_attr "mode" "<MODE>")]) @@ -3680,7 +3698,7 @@ (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3701,7 +3719,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3722,7 +3740,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c index 2e0e12aa045..2eef9e1e1a8 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c @@ -1,4 +1,4 @@ -/* { dg-do compile } */ +/* { dg-do compile { target { ! riscv_xtheadvector } } } */ /* { dg-skip-if "test rvv intrinsic" { *-*-* } { "*" } { "-march=rv*v*" } } */ void foo0 () {__rvv_bool64_t t;} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c index 3d81b179235..ef329e30785 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c @@ -1,4 +1,4 @@ /* { dg-do compile } */ /* { dg-options "-O3 -march=rv32gc -mabi=ilp32d" } */ -#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' extension enabled} } */ +#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' or 'XTHEADVECTOR' extension enabled} } */ diff --git a/gcc/testsuite/lib/target-supports.exp b/gcc/testsuite/lib/target-supports.exp index 7f13ff0ca56..70df6b1401c 100644 --- a/gcc/testsuite/lib/target-supports.exp +++ b/gcc/testsuite/lib/target-supports.exp @@ -1952,6 +1952,18 @@ proc check_effective_target_riscv_zbb { } { }] } +# Return 1 if the target arch supports the XTheadVector extension, 0 otherwise. +# Cache the result. + +proc check_effective_target_riscv_xtheadvector { } { + return [check_no_compiler_messages riscv_ext_xtheadvector assembly { + #ifndef __riscv_xtheadvector + #error "Not __riscv_xtheadvector" + #endif + }] +} + + # Return 1 if we can execute code when using dg-add-options riscv_v proc check_effective_target_riscv_v_ok { } { -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: Re：[PATCH v4] RISC-V: Handle differences between XTheadvector and Vector 2024-01-02 3:03 ` Re：[PATCH " joshua @ 2024-01-02 3:10 ` juzhe.zhong 2024-01-02 3:40 ` Re：Re：[PATCH " joshua 0 siblings, 1 reply; 133+ messages in thread From: juzhe.zhong @ 2024-01-02 3:10 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner [-- Attachment #1: Type: text/plain, Size: 72174 bytes --] Like riscv_output_move if (TARGET_THEADVECTOR) return vsetvli....no tail policy and mask policy. else return .... juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-02 11:03 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner 主题： Re：[PATCH v4] RISC-V: Handle differences between XTheadvector and Vector For vsetvl issues, what we want to do here is to directly remove "t" and "m". If we add TARGET_XTHEADVECTOR logic in case "p" in riscv_print_operand, how can we remove "t" and "m"? If I use "break", assembly like "th.vsetvli zero,a5,e8,m1,t,m" will be returned. if (TARGET_THEADVECTOR) ... else else if (code == CONST_INT) { /* Tail && Mask policy. */ asm_fprintf (file, "%s", IS_AGNOSTIC (UINTVAL (op)) ? "a" : "u"); } ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月2日(星期二) 10:00 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v4] RISC-V: Handle differences between XTheadvector and Vector + if (TARGET_XTHEADVECTOR) + return false; Move it to : if (TARGET_VECTOR && stringop_strategy & STRATEGY_VECTOR) { bool ok = riscv_vector::expand_block_move (dest, src, length); if (ok) return true; } (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (match_test "!TARGET_XTHEADVECTOR || rtx_equal_p (op, const0_rtx)") + (match_operand 0 "const_csr_operand")))) It's hard to trace. Change it into : (ior 1. TARGET_THEADVECTOR && rtx_equal_p (op, const0_rtx) 2. !TAGEET_THEADVECTOR && const_csr_operand) + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (mode, dest, src, + RVV_VLMAX, GEN_INT(VLMAX))); + return true; + } Move it outside legitimize_move It should be it: if (TARGET_THEADVECTOR) emit_th_move... DONE; else if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; vsetvli issues: I wonder whether we can use ASM_OUTPUT_OPCODE to recognize "ta,ma"/"ta,mu"/"tu,ma"/"tu,mu" and replace these 4 variants by "". So that we don't have tail policy and mask policy in vsetvli ASM string. Another alternative approach is we can change vsetlvi ASM rule: "vset%i1vli\t%0,%1,e%2,%m3,t%p4,m%p5" if (TARGET_THEADVECTOR) ... else else if (code == CONST_INT) { /* Tail && Mask policy. */ asm_fprintf (file, "%s", IS_AGNOSTIC (UINTVAL (op)) ? "a" : "u"); } in riscv.cc. The benefit is that we can avoid adding all th_vsetvl patterns and invasive code changs in VSETVL PASS. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2023-12-29 12:21 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v4] RISC-V: Handle differences between XTheadvector and Vector This patch is to handle the differences in instruction generation between Vector and XTheadVector. In this version, we only support partial xtheadvector instructions that leverage directly from current RVV1.0 with simple adding "th." prefix. For different name xtheadvector instructions but share same patterns as RVV1.0 instructions, we will use ASM targethook to rewrite the whole string of the instructions in the following patches. For some vector patterns that cannot be avoided, we use "!TARGET_XTHEADVECTOR" to disable them in vector.md in order not to generate instructions that xtheadvector does not support, like vmv1r and vsext.vf2. gcc/ChangeLog: * config.gcc: Add files for XTheadVector intrinsics. * config/riscv/autovec.md: Guard XTheadVector. * config/riscv/riscv-string.cc (expand_block_move): Guard XTheadVector. * config/riscv/riscv-v.cc (legitimize_move): New expansion. (get_prefer_tail_policy): Give specific value for tail. (get_prefer_mask_policy): Give specific value for mask. (vls_mode_valid_p): Avoid autovec. * config/riscv/riscv-vector-builtins-shapes.cc (check_type): (build_one): New function. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): (DEF_THEAD_RVV_FUNCTION): Add new marcos. (check_required_extensions): (handle_pragma_vector): * config/riscv/riscv-vector-builtins.h (RVV_REQUIRE_VECTOR): (RVV_REQUIRE_XTHEADVECTOR): Add RVV_REQUIRE_VECTOR and RVV_REQUIRE_XTHEADVECTOR. (struct function_group_info): * config/riscv/riscv-vector-switch.def (ENTRY): Disable fractional mode for the XTheadVector extension. (TUPLE_ENTRY): Likewise. * config/riscv/riscv-vsetvl.cc: Add functions for xtheadvector. * config/riscv/riscv.cc (riscv_v_ext_vls_mode_p): Guard XTheadVector. (riscv_v_adjust_bytesize): Likewise. (riscv_preferred_simd_mode): Likewsie. (riscv_autovectorize_vector_modes): Likewise. (riscv_vector_mode_supported_any_target_p): Likewise. (TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P): Likewise. * config/riscv/vector-iterators.md: Remove fractional LMUL. * config/riscv/vector.md: Include thead-vector.md. * config/riscv/riscv_th_vector.h: New file. * config/riscv/thead-vector.md: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/base/pragma-1.c: Add XTheadVector. * gcc.target/riscv/rvv/base/abi-1.c: Exclude XTheadVector. * lib/target-supports.exp: Add target for XTheadVector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config.gcc | 2 +- gcc/config/riscv/autovec.md | 2 +- gcc/config/riscv/predicates.md | 3 +- gcc/config/riscv/riscv-string.cc | 3 + gcc/config/riscv/riscv-v.cc | 13 +- .../riscv/riscv-vector-builtins-bases.cc | 3 + .../riscv/riscv-vector-builtins-shapes.cc | 23 +++ gcc/config/riscv/riscv-vector-switch.def | 150 +++++++------- gcc/config/riscv/riscv-vsetvl.cc | 10 + gcc/config/riscv/riscv.cc | 20 +- gcc/config/riscv/riscv_th_vector.h | 49 +++++ gcc/config/riscv/thead-vector.md | 142 +++++++++++++ gcc/config/riscv/vector-iterators.md | 186 +++++++++--------- gcc/config/riscv/vector.md | 36 +++- .../gcc.target/riscv/rvv/base/abi-1.c | 2 +- .../gcc.target/riscv/rvv/base/pragma-1.c | 2 +- gcc/testsuite/lib/target-supports.exp | 12 ++ 17 files changed, 471 insertions(+), 187 deletions(-) create mode 100644 gcc/config/riscv/riscv_th_vector.h create mode 100644 gcc/config/riscv/thead-vector.md diff --git a/gcc/config.gcc b/gcc/config.gcc index f0676c830e8..1445d98c147 100644 --- a/gcc/config.gcc +++ b/gcc/config.gcc @@ -549,7 +549,7 @@ riscv*) extra_objs="${extra_objs} riscv-vector-builtins.o riscv-vector-builtins-shapes.o riscv-vector-builtins-bases.o" extra_objs="${extra_objs} thead.o riscv-target-attr.o" d_target_objs="riscv-d.o" - extra_headers="riscv_vector.h" + extra_headers="riscv_vector.h riscv_th_vector.h" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.cc" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.h" ;; diff --git a/gcc/config/riscv/autovec.md b/gcc/config/riscv/autovec.md index 8b8a92f10a1..1fac56c7095 100644 --- a/gcc/config/riscv/autovec.md +++ b/gcc/config/riscv/autovec.md @@ -2579,7 +2579,7 @@ [(match_operand 0 "register_operand") (match_operand 1 "memory_operand") (match_operand:ANYI 2 "const_int_operand")] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" { riscv_vector::expand_rawmemchr(<MODE>mode, operands[0], operands[1], operands[2]); diff --git a/gcc/config/riscv/predicates.md b/gcc/config/riscv/predicates.md index b87a6900841..c5ab86d5db9 100644 --- a/gcc/config/riscv/predicates.md +++ b/gcc/config/riscv/predicates.md @@ -428,7 +428,8 @@ ;; Predicates for the V extension. (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (match_test "!TARGET_XTHEADVECTOR || rtx_equal_p (op, const0_rtx)") + (match_operand 0 "const_csr_operand")))) (define_special_predicate "autovec_length_operand" (ior (match_operand 0 "pmode_register_operand") diff --git a/gcc/config/riscv/riscv-string.cc b/gcc/config/riscv/riscv-string.cc index 11c1f74d0b3..ec8f3486fd8 100644 --- a/gcc/config/riscv/riscv-string.cc +++ b/gcc/config/riscv/riscv-string.cc @@ -808,6 +808,9 @@ expand_block_move (rtx dst_in, rtx src_in, rtx length_in) bnez a2, loop # Any more? ret # Return */ + if (TARGET_XTHEADVECTOR) + return false; + gcc_assert (TARGET_VECTOR); HOST_WIDE_INT potential_ew diff --git a/gcc/config/riscv/riscv-v.cc b/gcc/config/riscv/riscv-v.cc index 038ab084a37..5e9e45aecd2 100644 --- a/gcc/config/riscv/riscv-v.cc +++ b/gcc/config/riscv/riscv-v.cc @@ -1523,6 +1523,13 @@ legitimize_move (rtx dest, rtx *srcp) return true; } + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (mode, dest, src, + RVV_VLMAX, GEN_INT(VLMAX))); + return true; + } + if (riscv_v_ext_vls_mode_p (mode)) { if (GET_MODE_NUNITS (mode).to_constant () <= 31) @@ -1772,7 +1779,7 @@ get_prefer_tail_policy () compiler pick up either agnostic or undisturbed. Maybe we will have a compile option like -mprefer=agnostic to set this value???. */ - return TAIL_ANY; + return TARGET_XTHEADVECTOR ? TAIL_AGNOSTIC : TAIL_ANY; } /* Get prefer mask policy. */ @@ -1783,7 +1790,7 @@ get_prefer_mask_policy () compiler pick up either agnostic or undisturbed. Maybe we will have a compile option like -mprefer=agnostic to set this value???. */ - return MASK_ANY; + return TARGET_XTHEADVECTOR ? MASK_UNDISTURBED : MASK_ANY; } /* Get avl_type rtx. */ @@ -4383,7 +4390,7 @@ cmp_lmul_gt_one (machine_mode mode) bool vls_mode_valid_p (machine_mode vls_mode) { - if (!TARGET_VECTOR) + if (!TARGET_VECTOR || TARGET_XTHEADVECTOR) return false; if (riscv_autovec_preference == RVV_SCALABLE) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index c51affde353..2918c07ebf3 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -133,6 +133,9 @@ public: = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + if (TARGET_XTHEADVECTOR) + return e.generate_insn (code_for_th_vsetvl_no_side_effects (Pmode)); + /* TAIL_ANY. */ e.add_input_operand (Pmode, gen_int_mode (get_prefer_tail_policy (), Pmode)); diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 4a754e0228f..6b49404a1fa 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -33,6 +33,25 @@ namespace riscv_vector { +/* Check whether the RETURN_TYPE and ARGUMENT_TYPES are + valid for the function. */ + +static bool +check_type (tree return_type, vec<tree> &argument_types) +{ + tree arg; + unsigned i; + + if (!return_type) + return false; + + FOR_EACH_VEC_ELT (argument_types, i, arg) + if (!arg) + return false; + + return true; +} + /* Add one function instance for GROUP, using operand suffix at index OI, mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ static void @@ -49,6 +68,10 @@ build_one (function_builder &b, const function_group_info &group, group.ops_infos.types[vec_type_idx].index); b.allocate_argument_types (function_instance, argument_types); b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + b.add_overloaded_function (function_instance, *group.shape); b.add_unique_function (function_instance, (*group.shape), return_type, argument_types); diff --git a/gcc/config/riscv/riscv-vector-switch.def b/gcc/config/riscv/riscv-vector-switch.def index 5c9f9bcbc3e..f7a66b34bae 100644 --- a/gcc/config/riscv/riscv-vector-switch.def +++ b/gcc/config/riscv/riscv-vector-switch.def @@ -68,9 +68,9 @@ Encode the ratio of SEW/LMUL into the mask types. #endif /* Disable modes if TARGET_MIN_VLEN == 32. */ -ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) -ENTRY (RVVMF32BI, true, LMUL_F4, 32) -ENTRY (RVVMF16BI, true, LMUL_F2, 16) +ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F8, 64) +ENTRY (RVVMF32BI, true, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F4, 32) +ENTRY (RVVMF16BI, true, TARGET_XTHEADVECTOR ? LMUL_1 : LMUL_F2 , 16) ENTRY (RVVMF8BI, true, LMUL_1, 8) ENTRY (RVVMF4BI, true, LMUL_2, 4) ENTRY (RVVMF2BI, true, LMUL_4, 2) @@ -81,39 +81,39 @@ ENTRY (RVVM8QI, true, LMUL_8, 1) ENTRY (RVVM4QI, true, LMUL_4, 2) ENTRY (RVVM2QI, true, LMUL_2, 4) ENTRY (RVVM1QI, true, LMUL_1, 8) -ENTRY (RVVMF2QI, true, LMUL_F2, 16) -ENTRY (RVVMF4QI, true, LMUL_F4, 32) -ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) +ENTRY (RVVMF2QI, !TARGET_XTHEADVECTOR, LMUL_F2, 16) +ENTRY (RVVMF4QI, !TARGET_XTHEADVECTOR, LMUL_F4, 32) +ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F8, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8HI, true, LMUL_8, 2) ENTRY (RVVM4HI, true, LMUL_4, 4) ENTRY (RVVM2HI, true, LMUL_2, 8) ENTRY (RVVM1HI, true, LMUL_1, 16) -ENTRY (RVVMF2HI, true, LMUL_F2, 32) -ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HI, !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_16. */ ENTRY (RVVM8HF, TARGET_VECTOR_ELEN_FP_16, LMUL_8, 2) ENTRY (RVVM4HF, TARGET_VECTOR_ELEN_FP_16, LMUL_4, 4) ENTRY (RVVM2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_2, 8) ENTRY (RVVM1HF, TARGET_VECTOR_ELEN_FP_16, LMUL_1, 16) -ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_F2, 32) -ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8SI, true, LMUL_8, 4) ENTRY (RVVM4SI, true, LMUL_4, 8) ENTRY (RVVM2SI, true, LMUL_2, 16) ENTRY (RVVM1SI, true, LMUL_1, 32) -ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_32. */ ENTRY (RVVM8SF, TARGET_VECTOR_ELEN_FP_32, LMUL_8, 4) ENTRY (RVVM4SF, TARGET_VECTOR_ELEN_FP_32, LMUL_4, 8) ENTRY (RVVM2SF, TARGET_VECTOR_ELEN_FP_32, LMUL_2, 16) ENTRY (RVVM1SF, TARGET_VECTOR_ELEN_FP_32, LMUL_1, 32) -ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if !TARGET_VECTOR_ELEN_64. */ ENTRY (RVVM8DI, TARGET_VECTOR_ELEN_64, LMUL_8, 8) @@ -140,127 +140,127 @@ ENTRY (RVVM1DF, TARGET_VECTOR_ELEN_FP_64, LMUL_1, 64) #endif TUPLE_ENTRY (RVVM1x8QI, true, RVVM1QI, 8, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x8QI, true, RVVMF2QI, 8, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x8QI, true, RVVMF4QI, 8, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 8, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x8QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 8, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x8QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 8, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 8, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x7QI, true, RVVM1QI, 7, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x7QI, true, RVVMF2QI, 7, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x7QI, true, RVVMF4QI, 7, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 7, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x7QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 7, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x7QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 7, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 7, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x6QI, true, RVVM1QI, 6, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x6QI, true, RVVMF2QI, 6, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x6QI, true, RVVMF4QI, 6, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 6, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x6QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 6, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x6QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 6, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 6, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x5QI, true, RVVM1QI, 5, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x5QI, true, RVVMF2QI, 5, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x5QI, true, RVVMF4QI, 5, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 5, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x5QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 5, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x5QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 5, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 5, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x4QI, true, RVVM2QI, 4, LMUL_2, 4) TUPLE_ENTRY (RVVM1x4QI, true, RVVM1QI, 4, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x4QI, true, RVVMF2QI, 4, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x4QI, true, RVVMF4QI, 4, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 4, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x4QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 4, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x4QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 4, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 4, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x3QI, true, RVVM2QI, 3, LMUL_2, 4) TUPLE_ENTRY (RVVM1x3QI, true, RVVM1QI, 3, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x3QI, true, RVVMF2QI, 3, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x3QI, true, RVVMF4QI, 3, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 3, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x3QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 3, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x3QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 3, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 3, LMUL_F8, 64) TUPLE_ENTRY (RVVM4x2QI, true, RVVM4QI, 2, LMUL_4, 2) TUPLE_ENTRY (RVVM2x2QI, true, RVVM2QI, 2, LMUL_2, 4) TUPLE_ENTRY (RVVM1x2QI, true, RVVM1QI, 2, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x2QI, true, RVVMF2QI, 2, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x2QI, true, RVVMF4QI, 2, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 2, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x2QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 2, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x2QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 2, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 2, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x8HI, true, RVVM1HI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HI, true, RVVMF2HI, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HI, true, RVVM1HI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HI, true, RVVMF2HI, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HI, true, RVVM1HI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HI, true, RVVMF2HI, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HI, true, RVVM1HI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HI, true, RVVMF2HI, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HI, true, RVVM2HI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HI, true, RVVM1HI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HI, true, RVVMF2HI, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HI, true, RVVM2HI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HI, true, RVVM1HI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HI, true, RVVMF2HI, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HI, true, RVVM4HI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HI, true, RVVM2HI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HI, true, RVVM1HI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HI, true, RVVMF2HI, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM4HF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8SI, true, RVVM1SI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SI, true, RVVM1SI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SI, true, RVVM1SI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SI, true, RVVM1SI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SI, true, RVVM2SI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SI, true, RVVM1SI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SI, true, RVVM2SI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SI, true, RVVM1SI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SI, true, RVVM4SI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SI, true, RVVM2SI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SI, true, RVVM1SI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM4SF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 8, LMUL_1, 16) TUPLE_ENTRY (RVVM1x7DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 7, LMUL_1, 16) diff --git a/gcc/config/riscv/riscv-vsetvl.cc b/gcc/config/riscv/riscv-vsetvl.cc index eabaef80f89..c726253c107 100644 --- a/gcc/config/riscv/riscv-vsetvl.cc +++ b/gcc/config/riscv/riscv-vsetvl.cc @@ -1117,6 +1117,16 @@ public: avl = GEN_INT (0); rtx sew = gen_int_mode (get_sew (), Pmode); rtx vlmul = gen_int_mode (get_vlmul (), Pmode); + + if (TARGET_XTHEADVECTOR) { + if (change_vtype_only_p ()) + return gen_th_vsetvl_vtype_change_only (sew, vlmul); + else if (has_vl () && !ignore_vl) + return gen_th_vsetvl (Pmode, get_vl (), avl, sew, vlmul); + else + return gen_th_vsetvl_discard_result (Pmode, avl, sew, vlmul); + } + rtx ta = gen_int_mode (get_ta (), Pmode); rtx ma = gen_int_mode (get_ma (), Pmode); diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index ea1d59d9cf2..a80bf8d1a74 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -1406,6 +1406,9 @@ riscv_v_adjust_bytesize (machine_mode mode, int scale) { if (riscv_v_ext_vector_mode_p (mode)) { + if (TARGET_XTHEADVECTOR) + return BYTES_PER_RISCV_VECTOR; + poly_int64 nunits = GET_MODE_NUNITS (mode); poly_int64 mode_size = GET_MODE_SIZE (mode); @@ -9965,7 +9968,7 @@ riscv_use_divmod_expander (void) static machine_mode riscv_preferred_simd_mode (scalar_mode mode) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::preferred_simd_mode (mode); return word_mode; @@ -10316,7 +10319,7 @@ riscv_mode_priority (int, int n) unsigned int riscv_autovectorize_vector_modes (vector_modes *modes, bool all) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::autovectorize_vector_modes (modes, all); return default_autovectorize_vector_modes (modes, all); @@ -10499,6 +10502,16 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) return false; } +/* Implements target hook vector_mode_supported_any_target_p. */ + +static bool +riscv_vector_mode_supported_any_target_p (machine_mode mode) +{ + if (TARGET_XTHEADVECTOR) + return false; + return true; +} + /* Initialize the GCC target structure. */ #undef TARGET_ASM_ALIGNED_HI_OP #define TARGET_ASM_ALIGNED_HI_OP "\t.half\t" @@ -10842,6 +10855,9 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) #undef TARGET_PREFERRED_ELSE_VALUE #define TARGET_PREFERRED_ELSE_VALUE riscv_preferred_else_value +#undef TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P +#define TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P riscv_vector_mode_supported_any_target_p + struct gcc_target targetm = TARGET_INITIALIZER; #include "gt-riscv.h" diff --git a/gcc/config/riscv/riscv_th_vector.h b/gcc/config/riscv/riscv_th_vector.h new file mode 100644 index 00000000000..6f47e0c90a4 --- /dev/null +++ b/gcc/config/riscv/riscv_th_vector.h @@ -0,0 +1,49 @@ +/* RISC-V 'XTheadVector' Extension intrinsics include file. + Copyright (C) 2022-2023 Free Software Foundation, Inc. + + This file is part of GCC. + + GCC is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published + by the Free Software Foundation; either version 3, or (at your + option) any later version. + + GCC is distributed in the hope that it will be useful, but WITHOUT + ANY WARRANTY; without even the implied warranty of MERCHANTABILITY + or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public + License for more details. + + Under Section 7 of GPL version 3, you are granted additional + permissions described in the GCC Runtime Library Exception, version + 3.1, as published by the Free Software Foundation. + + You should have received a copy of the GNU General Public License and + a copy of the GCC Runtime Library Exception along with this program; + see the files COPYING3 and COPYING.RUNTIME respectively. If not, see + <" target="_blank">http://www.gnu.org/licenses/>. */ + +#ifndef __RISCV_TH_VECTOR_H +#define __RISCV_TH_VECTOR_H + +#include <stdint.h> +#include <stddef.h> + +#ifndef __riscv_xtheadvector +#error "XTheadVector intrinsics require the xtheadvector extension." +#else +#ifdef __cplusplus +extern "C" { +#endif + +/* NOTE: This implementation of riscv_th_vector.h is intentionally short. It does + not define the RVV types and intrinsic functions directly in C and C++ + code, but instead uses the following pragma to tell GCC to insert the + necessary type and function definitions itself. The net effect is the + same, and the file is a complete implementation of riscv_th_vector.h. */ +#pragma riscv intrinsic "vector" + +#ifdef __cplusplus +} +#endif // __cplusplus +#endif // __riscv_xtheadvector +#endif // __RISCV_TH_ECTOR_H diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md new file mode 100644 index 00000000000..f0aa196d93b --- /dev/null +++ b/gcc/config/riscv/thead-vector.md @@ -0,0 +1,142 @@ +(define_c_enum "unspec" [ + UNSPEC_TH_VWLDST +]) + +(define_mode_iterator V_VLS_VT [V VLS VT]) +(define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) + +(define_split + [(set (match_operand:V_VB_VLS_VT 0 "reg_or_mem_operand") + (match_operand:V_VB_VLS_VT 1 "reg_or_mem_operand"))] + "TARGET_XTHEADVECTOR" + [(const_int 0)] + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + }) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:V_VLS_VT 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:V_VLS_VT + [(match_operand:V_VLS_VT 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2")]) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:VB 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:VB + [(match_operand:VB 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2") + (set (attr "sew") (const_int 8)) + (set (attr "vlmul") (symbol_ref "riscv_vector::LMUL_1"))]) + +(define_insn "@th_vsetvl<mode>" + [(set (match_operand:P 0 "register_operand" "=r") + (unspec:P [(match_operand:P 1 "vector_length_operand" "rK") + (match_operand 2 "const_int_operand" "i") + (match_operand 3 "const_int_operand" "i")] UNSPEC_VSETVL)) + (set (reg:SI VL_REGNUM) + (unspec:SI [(match_dup 1) + (match_dup 2) + (match_dup 3)] UNSPEC_VSETVL)) + (set (reg:SI VTYPE_REGNUM) + (unspec:SI [(match_dup 2) + (match_dup 3)] UNSPEC_VSETVL))] + "TARGET_XTHEADVECTOR" + "vsetvli\t%0,%1,e%2,%m3" + [(set_attr "type" "vsetvl") + (set_attr "mode" "<MODE>") + (set (attr "sew") (symbol_ref "INTVAL (operands[2])")) + (set (attr "vlmul") (symbol_ref "INTVAL (operands[3])"))]) + +;; vsetvl zero,zero,vtype instruction. +;; This pattern has no side effects and does not set X0 register. +(define_insn "th_vsetvl_vtype_change_only" + [(set (reg:SI VTYPE_REGNUM) + (unspec:SI + [(match_operand 0 "const_int_operand" "i") + (match_operand 1 "const_int_operand" "i")] UNSPEC_VSETVL))] + "TARGET_XTHEADVECTOR" + "vsetvli\tzero,zero,e%0,%m1" + [(set_attr "type" "vsetvl") + (set_attr "mode" "SI") + (set (attr "sew") (symbol_ref "INTVAL (operands[0])")) + (set (attr "vlmul") (symbol_ref "INTVAL (operands[1])"))]) + +;; vsetvl zero,rs1,vtype instruction. +;; The reason we need this pattern since we should avoid setting X0 register +;; in vsetvl instruction pattern. +(define_insn "@th_vsetvl_discard_result<mode>" + [(set (reg:SI VL_REGNUM) + (unspec:SI [(match_operand:P 0 "vector_length_operand" "rK") + (match_operand 1 "const_int_operand" "i") + (match_operand 2 "const_int_operand" "i")] UNSPEC_VSETVL)) + (set (reg:SI VTYPE_REGNUM) + (unspec:SI [(match_dup 1) + (match_dup 2)] UNSPEC_VSETVL))] + "TARGET_XTHEADVECTOR" + "vsetvli\tzero,%0,e%1,%m2" + [(set_attr "type" "vsetvl") + (set_attr "mode" "<MODE>") + (set (attr "sew") (symbol_ref "INTVAL (operands[1])")) + (set (attr "vlmul") (symbol_ref "INTVAL (operands[2])"))]) + +;; It's emit by vsetvl/vsetvlmax intrinsics with no side effects. +;; Since we have many optmization passes from "expand" to "reload_completed", +;; such pattern can allow us gain benefits of these optimizations. +(define_insn_and_split "@th_vsetvl<mode>_no_side_effects" + [(set (match_operand:P 0 "register_operand" "=r") + (unspec:P [(match_operand:P 1 "vector_length_operand" "rK") + (match_operand 2 "const_int_operand" "i") + (match_operand 3 "const_int_operand" "i")] UNSPEC_VSETVL))] + "TARGET_XTHEADVECTOR" + "#" + "&& epilogue_completed" + [(parallel + [(set (match_dup 0) + (unspec:P [(match_dup 1) (match_dup 2) (match_dup 3)] UNSPEC_VSETVL)) + (set (reg:SI VL_REGNUM) + (unspec:SI [(match_dup 1) (match_dup 2) (match_dup 3)] UNSPEC_VSETVL)) + (set (reg:SI VTYPE_REGNUM) + (unspec:SI [(match_dup 2) (match_dup 3)] UNSPEC_VSETVL))])] + "" + [(set_attr "type" "vsetvl") + (set_attr "mode" "SI")]) diff --git a/gcc/config/riscv/vector-iterators.md b/gcc/config/riscv/vector-iterators.md index 5f5f7b5b986..c0fc7a2441d 100644 --- a/gcc/config/riscv/vector-iterators.md +++ b/gcc/config/riscv/vector-iterators.md @@ -109,11 +109,11 @@ ]) (define_mode_iterator VI [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -128,11 +128,11 @@ ;; allow the instruction and mode to be matched during combine et al. (define_mode_iterator VF [ (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -140,11 +140,11 @@ (define_mode_iterator VF_ZVFHMIN [ (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -271,16 +271,16 @@ ]) (define_mode_iterator VEEWEXT2 [ - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -290,10 +290,10 @@ ]) (define_mode_iterator VEEWEXT4 [ - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -311,59 +311,59 @@ ]) (define_mode_iterator VEEWTRUNC2 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") (RVVM4SI "TARGET_64BIT") (RVVM2SI "TARGET_64BIT") (RVVM1SI "TARGET_64BIT") - (RVVMF2SI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEEWTRUNC4 [ - RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM2HI "TARGET_64BIT") (RVVM1HI "TARGET_64BIT") - (RVVMF2HI "TARGET_64BIT") - (RVVMF4HI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2HI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") (RVVM1HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEEWTRUNC8 [ (RVVM1QI "TARGET_64BIT") - (RVVMF2QI "TARGET_64BIT") - (RVVMF4QI "TARGET_64BIT") - (RVVMF8QI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2QI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF4QI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEI16 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -452,11 +452,11 @@ ]) (define_mode_iterator VFULLI [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_FULL_V") (RVVM4DI "TARGET_FULL_V") (RVVM2DI "TARGET_FULL_V") (RVVM1DI "TARGET_FULL_V") @@ -509,17 +509,17 @@ ]) (define_mode_iterator VI_QH [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") ]) (define_mode_iterator VI_QHS [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -560,11 +560,11 @@ ]) (define_mode_iterator VI_QHS_NO_M8 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -603,11 +603,11 @@ (define_mode_iterator VF_HS [ (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (V1HF "riscv_vector::vls_mode_valid_p (V1HFmode) && TARGET_ZVFH") (V2HF "riscv_vector::vls_mode_valid_p (V2HFmode) && TARGET_ZVFH") @@ -638,12 +638,12 @@ (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") - (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (V1HF "riscv_vector::vls_mode_valid_p (V1HFmode) && TARGET_ZVFH") (V2HF "riscv_vector::vls_mode_valid_p (V2HFmode) && TARGET_ZVFH") @@ -674,11 +674,11 @@ ]) (define_mode_iterator V_VLSI_QHS [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -756,27 +756,27 @@ ;; E.g. when index mode = RVVM8QImde and Pmode = SImode, if it is not zero_extend or ;; scalar != 1, such gather/scatter is not allowed since we don't have RVVM32SImode. (define_mode_iterator RATIO64 [ - (RVVMF8QI "TARGET_MIN_VLEN > 32") - (RVVMF4HI "TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM1DI "TARGET_VECTOR_ELEN_64") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") ]) (define_mode_iterator RATIO32 [ - RVVMF4QI - RVVMF2HI + (RVVMF4QI "!TARGET_XTHEADVECTOR") + (RVVMF2HI "!TARGET_XTHEADVECTOR") RVVM1SI (RVVM2DI "TARGET_VECTOR_ELEN_64") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") ]) (define_mode_iterator RATIO16 [ - RVVMF2QI + (RVVMF2QI "!TARGET_XTHEADVECTOR") RVVM1HI RVVM2SI (RVVM4DI "TARGET_VECTOR_ELEN_64") @@ -814,21 +814,21 @@ ]) (define_mode_iterator RATIO64I [ - (RVVMF8QI "TARGET_MIN_VLEN > 32") - (RVVMF4HI "TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM1DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") ]) (define_mode_iterator RATIO32I [ - RVVMF4QI - RVVMF2HI + (RVVMF4QI "!TARGET_XTHEADVECTOR") + (RVVMF2HI "!TARGET_XTHEADVECTOR") RVVM1SI (RVVM2DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") ]) (define_mode_iterator RATIO16I [ - RVVMF2QI + (RVVMF2QI "!TARGET_XTHEADVECTOR") RVVM1HI RVVM2SI (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") @@ -873,21 +873,21 @@ ]) (define_mode_iterator V_FRACT [ - RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") ]) (define_mode_iterator VWEXTI [ - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -933,7 +933,7 @@ (RVVM4SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -966,7 +966,7 @@ (RVVM4SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -996,7 +996,7 @@ (define_mode_iterator VWCONVERTI [ (RVVM8SI "TARGET_ZVFH") (RVVM4SI "TARGET_ZVFH") (RVVM2SI "TARGET_ZVFH") (RVVM1SI "TARGET_ZVFH") - (RVVMF2SI "TARGET_ZVFH") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_ZVFH") (RVVM8DI "TARGET_VECTOR_ELEN_64 && TARGET_VECTOR_ELEN_FP_32") (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_VECTOR_ELEN_FP_32") @@ -1045,7 +1045,7 @@ ]) (define_mode_iterator VQEXTI [ - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -1456,11 +1456,11 @@ ;; VINDEXED [VI8 VI16 VI32 (VI64 "TARGET_64BIT")]. (define_mode_iterator VINDEXED [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") @@ -1468,12 +1468,12 @@ (RVVM1DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64 && TARGET_64BIT") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64 && TARGET_64BIT") @@ -3173,11 +3173,11 @@ (define_mode_iterator V_VLS_F_CONVERT_SI [ (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") - (RVVMF2HF "TARGET_ZVFH") (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") @@ -3290,12 +3290,12 @@ ]) (define_mode_iterator V_VLS_F_CONVERT_DI [ - (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index b5a9055cdc4..9e7a487759a 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -83,7 +83,7 @@ ;; check. However, we need default value of SEW for vsetvl instruction since there ;; is no field for ratio in the vsetvl instruction encoding. (define_attr "sew" "" - (cond [(eq_attr "mode" "RVVMF64BI,RVVMF32BI,RVVMF16BI,RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ + (cond [(eq_attr "mode" "RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ RVVM8QI,RVVM4QI,RVVM2QI,RVVM1QI,RVVMF2QI,RVVMF4QI,RVVMF8QI,\ RVVM1x8QI,RVVMF2x8QI,RVVMF4x8QI,RVVMF8x8QI,\ RVVM1x7QI,RVVMF2x7QI,RVVMF4x7QI,RVVMF8x7QI,\ @@ -95,6 +95,18 @@ V1QI,V2QI,V4QI,V8QI,V16QI,V32QI,V64QI,V128QI,V256QI,V512QI,V1024QI,V2048QI,V4096QI,\ V1BI,V2BI,V4BI,V8BI,V16BI,V32BI,V64BI,V128BI,V256BI,V512BI,V1024BI,V2048BI,V4096BI") (const_int 8) + (eq_attr "mode" "RVVMF16BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 16) + (const_int 8)) + (eq_attr "mode" "RVVMF32BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 32) + (const_int 8)) + (eq_attr "mode" "RVVMF64BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 64) + (const_int 8)) (eq_attr "mode" "RVVM8HI,RVVM4HI,RVVM2HI,RVVM1HI,RVVMF2HI,RVVMF4HI,\ RVVM1x8HI,RVVMF2x8HI,RVVMF4x8HI,\ RVVM1x7HI,RVVMF2x7HI,RVVMF4x7HI,\ @@ -155,9 +167,9 @@ (eq_attr "mode" "RVVM4QI,RVVMF2BI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2QI,RVVMF4BI") (symbol_ref "riscv_vector::LMUL_2") (eq_attr "mode" "RVVM1QI,RVVMF8BI") (symbol_ref "riscv_vector::LMUL_1") - (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "riscv_vector::LMUL_F2") - (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "riscv_vector::LMUL_F4") - (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "riscv_vector::LMUL_F8") + (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F2") + (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F4") + (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F8") (eq_attr "mode" "RVVM8HI") (symbol_ref "riscv_vector::LMUL_8") (eq_attr "mode" "RVVM4HI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2HI") (symbol_ref "riscv_vector::LMUL_2") @@ -428,6 +440,10 @@ vislide1up,vislide1down,vfslide1up,vfslide1down,\ vgather,vcompress,vlsegdux,vlsegdox,vssegtux,vssegtox") (const_int INVALID_ATTRIBUTE) + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE) (eq_attr "mode" "RVVM8QI,RVVM1BI") (const_int 1) (eq_attr "mode" "RVVM4QI,RVVMF2BI") (const_int 2) (eq_attr "mode" "RVVM2QI,RVVMF4BI") (const_int 4) @@ -888,6 +904,8 @@ (symbol_ref "riscv_vector::FRM_DYN")] (symbol_ref "riscv_vector::FRM_NONE"))) +(include "thead-vector.md") + ;; ----------------------------------------------------------------- ;; ---- Miscellaneous Operations ;; ----------------------------------------------------------------- @@ -1097,7 +1115,7 @@ (define_insn "*mov<mode>_whole" [(set (match_operand:V_WHOLE 0 "reg_or_mem_operand" "=vr, m,vr") (match_operand:V_WHOLE 1 "reg_or_mem_operand" " m,vr,vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "@ vl%m1re<sew>.v\t%0,%1 vs%m1r.v\t%1,%0 @@ -1125,7 +1143,7 @@ (define_insn "*mov<mode>" [(set (match_operand:VB 0 "register_operand" "=vr") (match_operand:VB 1 "register_operand" " vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "vmv1r.v\t%0,%1" [(set_attr "type" "vmov") (set_attr "mode" "<MODE>")]) @@ -3680,7 +3698,7 @@ (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3701,7 +3719,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3722,7 +3740,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c index 2e0e12aa045..2eef9e1e1a8 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c @@ -1,4 +1,4 @@ -/* { dg-do compile } */ +/* { dg-do compile { target { ! riscv_xtheadvector } } } */ /* { dg-skip-if "test rvv intrinsic" { *-*-* } { "*" } { "-march=rv*v*" } } */ void foo0 () {__rvv_bool64_t t;} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c index 3d81b179235..ef329e30785 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c @@ -1,4 +1,4 @@ /* { dg-do compile } */ /* { dg-options "-O3 -march=rv32gc -mabi=ilp32d" } */ -#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' extension enabled} } */ +#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' or 'XTHEADVECTOR' extension enabled} } */ diff --git a/gcc/testsuite/lib/target-supports.exp b/gcc/testsuite/lib/target-supports.exp index 7f13ff0ca56..70df6b1401c 100644 --- a/gcc/testsuite/lib/target-supports.exp +++ b/gcc/testsuite/lib/target-supports.exp @@ -1952,6 +1952,18 @@ proc check_effective_target_riscv_zbb { } { }] } +# Return 1 if the target arch supports the XTheadVector extension, 0 otherwise. +# Cache the result. + +proc check_effective_target_riscv_xtheadvector { } { + return [check_no_compiler_messages riscv_ext_xtheadvector assembly { + #ifndef __riscv_xtheadvector + #error "Not __riscv_xtheadvector" + #endif + }] +} + + # Return 1 if we can execute code when using dg-add-options riscv_v proc check_effective_target_riscv_v_ok { } { -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re：Re：[PATCH v4] RISC-V: Handle differences between XTheadvector and Vector 2024-01-02 3:10 ` juzhe.zhong @ 2024-01-02 3:40 ` joshua 2024-01-02 3:41 ` Re：[PATCH " juzhe.zhong 0 siblings, 1 reply; 133+ messages in thread From: joshua @ 2024-01-02 3:40 UTC (permalink / raw) To: juzhe.zhong, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner But the riscv_print_operand() function returns void. We cannot return instructions like riscv_output_move. I think the briefest approach is to add some logic in the vsetvl patterns. "TARGET_VECTOR" { return TARGET_XTHEADVECTOR ? "vsetvli\t%0,%1,e%2,%m3" : "vset%i1vli\t%0,%1,e%2,%m3,t%p4,m%p5"; } Only 3 patterns need to be modified and I don't think it is too invasive. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月2日(星期二) 11:10 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu> 主 题：Re: Re：[PATCH v4] RISC-V: Handle differences between XTheadvector and Vector Like riscv_output_move if (TARGET_THEADVECTOR) return vsetvli....no tail policy and mask policy. else return .... juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-02 11:03 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner 主题： Re：[PATCH v4] RISC-V: Handle differences between XTheadvector and Vector For vsetvl issues, what we want to do here is to directly remove "t" and "m". If we add TARGET_XTHEADVECTOR logic in case "p" in riscv_print_operand, how can we remove "t" and "m"? If I use "break", assembly like "th.vsetvli zero,a5,e8,m1,t,m" will be returned. if (TARGET_THEADVECTOR) ... else else if (code == CONST_INT) { /* Tail && Mask policy. */ asm_fprintf (file, "%s", IS_AGNOSTIC (UINTVAL (op)) ? "a" : "u"); } ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月2日(星期二) 10:00 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v4] RISC-V: Handle differences between XTheadvector and Vector + if (TARGET_XTHEADVECTOR) + return false; Move it to : if (TARGET_VECTOR && stringop_strategy & STRATEGY_VECTOR) { bool ok = riscv_vector::expand_block_move (dest, src, length); if (ok) return true; } (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (match_test "!TARGET_XTHEADVECTOR || rtx_equal_p (op, const0_rtx)") + (match_operand 0 "const_csr_operand")))) It's hard to trace. Change it into : (ior 1. TARGET_THEADVECTOR && rtx_equal_p (op, const0_rtx) 2. !TAGEET_THEADVECTOR && const_csr_operand) + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (mode, dest, src, + RVV_VLMAX, GEN_INT(VLMAX))); + return true; + } Move it outside legitimize_move It should be it: if (TARGET_THEADVECTOR) emit_th_move... DONE; else if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; vsetvli issues: I wonder whether we can use ASM_OUTPUT_OPCODE to recognize "ta,ma"/"ta,mu"/"tu,ma"/"tu,mu" and replace these 4 variants by "". So that we don't have tail policy and mask policy in vsetvli ASM string. Another alternative approach is we can change vsetlvi ASM rule: "vset%i1vli\t%0,%1,e%2,%m3,t%p4,m%p5" if (TARGET_THEADVECTOR) ... else else if (code == CONST_INT) { /* Tail && Mask policy. */ asm_fprintf (file, "%s", IS_AGNOSTIC (UINTVAL (op)) ? "a" : "u"); } in riscv.cc. The benefit is that we can avoid adding all th_vsetvl patterns and invasive code changs in VSETVL PASS. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2023-12-29 12:21 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v4] RISC-V: Handle differences between XTheadvector and Vector This patch is to handle the differences in instruction generation between Vector and XTheadVector. In this version, we only support partial xtheadvector instructions that leverage directly from current RVV1.0 with simple adding "th." prefix. For different name xtheadvector instructions but share same patterns as RVV1.0 instructions, we will use ASM targethook to rewrite the whole string of the instructions in the following patches. For some vector patterns that cannot be avoided, we use "!TARGET_XTHEADVECTOR" to disable them in vector.md in order not to generate instructions that xtheadvector does not support, like vmv1r and vsext.vf2. gcc/ChangeLog: * config.gcc: Add files for XTheadVector intrinsics. * config/riscv/autovec.md: Guard XTheadVector. * config/riscv/riscv-string.cc (expand_block_move): Guard XTheadVector. * config/riscv/riscv-v.cc (legitimize_move): New expansion. (get_prefer_tail_policy): Give specific value for tail. (get_prefer_mask_policy): Give specific value for mask. (vls_mode_valid_p): Avoid autovec. * config/riscv/riscv-vector-builtins-shapes.cc (check_type): (build_one): New function. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): (DEF_THEAD_RVV_FUNCTION): Add new marcos. (check_required_extensions): (handle_pragma_vector): * config/riscv/riscv-vector-builtins.h (RVV_REQUIRE_VECTOR): (RVV_REQUIRE_XTHEADVECTOR): Add RVV_REQUIRE_VECTOR and RVV_REQUIRE_XTHEADVECTOR. (struct function_group_info): * config/riscv/riscv-vector-switch.def (ENTRY): Disable fractional mode for the XTheadVector extension. (TUPLE_ENTRY): Likewise. * config/riscv/riscv-vsetvl.cc: Add functions for xtheadvector. * config/riscv/riscv.cc (riscv_v_ext_vls_mode_p): Guard XTheadVector. (riscv_v_adjust_bytesize): Likewise. (riscv_preferred_simd_mode): Likewsie. (riscv_autovectorize_vector_modes): Likewise. (riscv_vector_mode_supported_any_target_p): Likewise. (TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P): Likewise. * config/riscv/vector-iterators.md: Remove fractional LMUL. * config/riscv/vector.md: Include thead-vector.md. * config/riscv/riscv_th_vector.h: New file. * config/riscv/thead-vector.md: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/base/pragma-1.c: Add XTheadVector. * gcc.target/riscv/rvv/base/abi-1.c: Exclude XTheadVector. * lib/target-supports.exp: Add target for XTheadVector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config.gcc | 2 +- gcc/config/riscv/autovec.md | 2 +- gcc/config/riscv/predicates.md | 3 +- gcc/config/riscv/riscv-string.cc | 3 + gcc/config/riscv/riscv-v.cc | 13 +- .../riscv/riscv-vector-builtins-bases.cc | 3 + .../riscv/riscv-vector-builtins-shapes.cc | 23 +++ gcc/config/riscv/riscv-vector-switch.def | 150 +++++++------- gcc/config/riscv/riscv-vsetvl.cc | 10 + gcc/config/riscv/riscv.cc | 20 +- gcc/config/riscv/riscv_th_vector.h | 49 +++++ gcc/config/riscv/thead-vector.md | 142 +++++++++++++ gcc/config/riscv/vector-iterators.md | 186 +++++++++--------- gcc/config/riscv/vector.md | 36 +++- .../gcc.target/riscv/rvv/base/abi-1.c | 2 +- .../gcc.target/riscv/rvv/base/pragma-1.c | 2 +- gcc/testsuite/lib/target-supports.exp | 12 ++ 17 files changed, 471 insertions(+), 187 deletions(-) create mode 100644 gcc/config/riscv/riscv_th_vector.h create mode 100644 gcc/config/riscv/thead-vector.md diff --git a/gcc/config.gcc b/gcc/config.gcc index f0676c830e8..1445d98c147 100644 --- a/gcc/config.gcc +++ b/gcc/config.gcc @@ -549,7 +549,7 @@ riscv*) extra_objs="${extra_objs} riscv-vector-builtins.o riscv-vector-builtins-shapes.o riscv-vector-builtins-bases.o" extra_objs="${extra_objs} thead.o riscv-target-attr.o" d_target_objs="riscv-d.o" - extra_headers="riscv_vector.h" + extra_headers="riscv_vector.h riscv_th_vector.h" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.cc" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.h" ;; diff --git a/gcc/config/riscv/autovec.md b/gcc/config/riscv/autovec.md index 8b8a92f10a1..1fac56c7095 100644 --- a/gcc/config/riscv/autovec.md +++ b/gcc/config/riscv/autovec.md @@ -2579,7 +2579,7 @@ [(match_operand 0 "register_operand") (match_operand 1 "memory_operand") (match_operand:ANYI 2 "const_int_operand")] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" { riscv_vector::expand_rawmemchr(<MODE>mode, operands[0], operands[1], operands[2]); diff --git a/gcc/config/riscv/predicates.md b/gcc/config/riscv/predicates.md index b87a6900841..c5ab86d5db9 100644 --- a/gcc/config/riscv/predicates.md +++ b/gcc/config/riscv/predicates.md @@ -428,7 +428,8 @@ ;; Predicates for the V extension. (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (match_test "!TARGET_XTHEADVECTOR || rtx_equal_p (op, const0_rtx)") + (match_operand 0 "const_csr_operand")))) (define_special_predicate "autovec_length_operand" (ior (match_operand 0 "pmode_register_operand") diff --git a/gcc/config/riscv/riscv-string.cc b/gcc/config/riscv/riscv-string.cc index 11c1f74d0b3..ec8f3486fd8 100644 --- a/gcc/config/riscv/riscv-string.cc +++ b/gcc/config/riscv/riscv-string.cc @@ -808,6 +808,9 @@ expand_block_move (rtx dst_in, rtx src_in, rtx length_in) bnez a2, loop # Any more? ret # Return */ + if (TARGET_XTHEADVECTOR) + return false; + gcc_assert (TARGET_VECTOR); HOST_WIDE_INT potential_ew diff --git a/gcc/config/riscv/riscv-v.cc b/gcc/config/riscv/riscv-v.cc index 038ab084a37..5e9e45aecd2 100644 --- a/gcc/config/riscv/riscv-v.cc +++ b/gcc/config/riscv/riscv-v.cc @@ -1523,6 +1523,13 @@ legitimize_move (rtx dest, rtx *srcp) return true; } + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (mode, dest, src, + RVV_VLMAX, GEN_INT(VLMAX))); + return true; + } + if (riscv_v_ext_vls_mode_p (mode)) { if (GET_MODE_NUNITS (mode).to_constant () <= 31) @@ -1772,7 +1779,7 @@ get_prefer_tail_policy () compiler pick up either agnostic or undisturbed. Maybe we will have a compile option like -mprefer=agnostic to set this value???. */ - return TAIL_ANY; + return TARGET_XTHEADVECTOR ? TAIL_AGNOSTIC : TAIL_ANY; } /* Get prefer mask policy. */ @@ -1783,7 +1790,7 @@ get_prefer_mask_policy () compiler pick up either agnostic or undisturbed. Maybe we will have a compile option like -mprefer=agnostic to set this value???. */ - return MASK_ANY; + return TARGET_XTHEADVECTOR ? MASK_UNDISTURBED : MASK_ANY; } /* Get avl_type rtx. */ @@ -4383,7 +4390,7 @@ cmp_lmul_gt_one (machine_mode mode) bool vls_mode_valid_p (machine_mode vls_mode) { - if (!TARGET_VECTOR) + if (!TARGET_VECTOR || TARGET_XTHEADVECTOR) return false; if (riscv_autovec_preference == RVV_SCALABLE) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index c51affde353..2918c07ebf3 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -133,6 +133,9 @@ public: = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + if (TARGET_XTHEADVECTOR) + return e.generate_insn (code_for_th_vsetvl_no_side_effects (Pmode)); + /* TAIL_ANY. */ e.add_input_operand (Pmode, gen_int_mode (get_prefer_tail_policy (), Pmode)); diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 4a754e0228f..6b49404a1fa 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -33,6 +33,25 @@ namespace riscv_vector { +/* Check whether the RETURN_TYPE and ARGUMENT_TYPES are + valid for the function. */ + +static bool +check_type (tree return_type, vec<tree> &argument_types) +{ + tree arg; + unsigned i; + + if (!return_type) + return false; + + FOR_EACH_VEC_ELT (argument_types, i, arg) + if (!arg) + return false; + + return true; +} + /* Add one function instance for GROUP, using operand suffix at index OI, mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ static void @@ -49,6 +68,10 @@ build_one (function_builder &b, const function_group_info &group, group.ops_infos.types[vec_type_idx].index); b.allocate_argument_types (function_instance, argument_types); b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + b.add_overloaded_function (function_instance, *group.shape); b.add_unique_function (function_instance, (*group.shape), return_type, argument_types); diff --git a/gcc/config/riscv/riscv-vector-switch.def b/gcc/config/riscv/riscv-vector-switch.def index 5c9f9bcbc3e..f7a66b34bae 100644 --- a/gcc/config/riscv/riscv-vector-switch.def +++ b/gcc/config/riscv/riscv-vector-switch.def @@ -68,9 +68,9 @@ Encode the ratio of SEW/LMUL into the mask types. #endif /* Disable modes if TARGET_MIN_VLEN == 32. */ -ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) -ENTRY (RVVMF32BI, true, LMUL_F4, 32) -ENTRY (RVVMF16BI, true, LMUL_F2, 16) +ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F8, 64) +ENTRY (RVVMF32BI, true, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F4, 32) +ENTRY (RVVMF16BI, true, TARGET_XTHEADVECTOR ? LMUL_1 : LMUL_F2 , 16) ENTRY (RVVMF8BI, true, LMUL_1, 8) ENTRY (RVVMF4BI, true, LMUL_2, 4) ENTRY (RVVMF2BI, true, LMUL_4, 2) @@ -81,39 +81,39 @@ ENTRY (RVVM8QI, true, LMUL_8, 1) ENTRY (RVVM4QI, true, LMUL_4, 2) ENTRY (RVVM2QI, true, LMUL_2, 4) ENTRY (RVVM1QI, true, LMUL_1, 8) -ENTRY (RVVMF2QI, true, LMUL_F2, 16) -ENTRY (RVVMF4QI, true, LMUL_F4, 32) -ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) +ENTRY (RVVMF2QI, !TARGET_XTHEADVECTOR, LMUL_F2, 16) +ENTRY (RVVMF4QI, !TARGET_XTHEADVECTOR, LMUL_F4, 32) +ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F8, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8HI, true, LMUL_8, 2) ENTRY (RVVM4HI, true, LMUL_4, 4) ENTRY (RVVM2HI, true, LMUL_2, 8) ENTRY (RVVM1HI, true, LMUL_1, 16) -ENTRY (RVVMF2HI, true, LMUL_F2, 32) -ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HI, !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_16. */ ENTRY (RVVM8HF, TARGET_VECTOR_ELEN_FP_16, LMUL_8, 2) ENTRY (RVVM4HF, TARGET_VECTOR_ELEN_FP_16, LMUL_4, 4) ENTRY (RVVM2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_2, 8) ENTRY (RVVM1HF, TARGET_VECTOR_ELEN_FP_16, LMUL_1, 16) -ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_F2, 32) -ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8SI, true, LMUL_8, 4) ENTRY (RVVM4SI, true, LMUL_4, 8) ENTRY (RVVM2SI, true, LMUL_2, 16) ENTRY (RVVM1SI, true, LMUL_1, 32) -ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_32. */ ENTRY (RVVM8SF, TARGET_VECTOR_ELEN_FP_32, LMUL_8, 4) ENTRY (RVVM4SF, TARGET_VECTOR_ELEN_FP_32, LMUL_4, 8) ENTRY (RVVM2SF, TARGET_VECTOR_ELEN_FP_32, LMUL_2, 16) ENTRY (RVVM1SF, TARGET_VECTOR_ELEN_FP_32, LMUL_1, 32) -ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if !TARGET_VECTOR_ELEN_64. */ ENTRY (RVVM8DI, TARGET_VECTOR_ELEN_64, LMUL_8, 8) @@ -140,127 +140,127 @@ ENTRY (RVVM1DF, TARGET_VECTOR_ELEN_FP_64, LMUL_1, 64) #endif TUPLE_ENTRY (RVVM1x8QI, true, RVVM1QI, 8, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x8QI, true, RVVMF2QI, 8, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x8QI, true, RVVMF4QI, 8, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 8, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x8QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 8, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x8QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 8, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 8, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x7QI, true, RVVM1QI, 7, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x7QI, true, RVVMF2QI, 7, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x7QI, true, RVVMF4QI, 7, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 7, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x7QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 7, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x7QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 7, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 7, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x6QI, true, RVVM1QI, 6, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x6QI, true, RVVMF2QI, 6, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x6QI, true, RVVMF4QI, 6, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 6, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x6QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 6, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x6QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 6, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 6, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x5QI, true, RVVM1QI, 5, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x5QI, true, RVVMF2QI, 5, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x5QI, true, RVVMF4QI, 5, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 5, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x5QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 5, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x5QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 5, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 5, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x4QI, true, RVVM2QI, 4, LMUL_2, 4) TUPLE_ENTRY (RVVM1x4QI, true, RVVM1QI, 4, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x4QI, true, RVVMF2QI, 4, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x4QI, true, RVVMF4QI, 4, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 4, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x4QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 4, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x4QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 4, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 4, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x3QI, true, RVVM2QI, 3, LMUL_2, 4) TUPLE_ENTRY (RVVM1x3QI, true, RVVM1QI, 3, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x3QI, true, RVVMF2QI, 3, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x3QI, true, RVVMF4QI, 3, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 3, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x3QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 3, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x3QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 3, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 3, LMUL_F8, 64) TUPLE_ENTRY (RVVM4x2QI, true, RVVM4QI, 2, LMUL_4, 2) TUPLE_ENTRY (RVVM2x2QI, true, RVVM2QI, 2, LMUL_2, 4) TUPLE_ENTRY (RVVM1x2QI, true, RVVM1QI, 2, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x2QI, true, RVVMF2QI, 2, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x2QI, true, RVVMF4QI, 2, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 2, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x2QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 2, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x2QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 2, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 2, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x8HI, true, RVVM1HI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HI, true, RVVMF2HI, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HI, true, RVVM1HI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HI, true, RVVMF2HI, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HI, true, RVVM1HI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HI, true, RVVMF2HI, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HI, true, RVVM1HI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HI, true, RVVMF2HI, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HI, true, RVVM2HI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HI, true, RVVM1HI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HI, true, RVVMF2HI, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HI, true, RVVM2HI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HI, true, RVVM1HI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HI, true, RVVMF2HI, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HI, true, RVVM4HI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HI, true, RVVM2HI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HI, true, RVVM1HI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HI, true, RVVMF2HI, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM4HF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8SI, true, RVVM1SI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SI, true, RVVM1SI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SI, true, RVVM1SI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SI, true, RVVM1SI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SI, true, RVVM2SI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SI, true, RVVM1SI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SI, true, RVVM2SI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SI, true, RVVM1SI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SI, true, RVVM4SI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SI, true, RVVM2SI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SI, true, RVVM1SI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM4SF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 8, LMUL_1, 16) TUPLE_ENTRY (RVVM1x7DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 7, LMUL_1, 16) diff --git a/gcc/config/riscv/riscv-vsetvl.cc b/gcc/config/riscv/riscv-vsetvl.cc index eabaef80f89..c726253c107 100644 --- a/gcc/config/riscv/riscv-vsetvl.cc +++ b/gcc/config/riscv/riscv-vsetvl.cc @@ -1117,6 +1117,16 @@ public: avl = GEN_INT (0); rtx sew = gen_int_mode (get_sew (), Pmode); rtx vlmul = gen_int_mode (get_vlmul (), Pmode); + + if (TARGET_XTHEADVECTOR) { + if (change_vtype_only_p ()) + return gen_th_vsetvl_vtype_change_only (sew, vlmul); + else if (has_vl () && !ignore_vl) + return gen_th_vsetvl (Pmode, get_vl (), avl, sew, vlmul); + else + return gen_th_vsetvl_discard_result (Pmode, avl, sew, vlmul); + } + rtx ta = gen_int_mode (get_ta (), Pmode); rtx ma = gen_int_mode (get_ma (), Pmode); diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index ea1d59d9cf2..a80bf8d1a74 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -1406,6 +1406,9 @@ riscv_v_adjust_bytesize (machine_mode mode, int scale) { if (riscv_v_ext_vector_mode_p (mode)) { + if (TARGET_XTHEADVECTOR) + return BYTES_PER_RISCV_VECTOR; + poly_int64 nunits = GET_MODE_NUNITS (mode); poly_int64 mode_size = GET_MODE_SIZE (mode); @@ -9965,7 +9968,7 @@ riscv_use_divmod_expander (void) static machine_mode riscv_preferred_simd_mode (scalar_mode mode) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::preferred_simd_mode (mode); return word_mode; @@ -10316,7 +10319,7 @@ riscv_mode_priority (int, int n) unsigned int riscv_autovectorize_vector_modes (vector_modes *modes, bool all) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::autovectorize_vector_modes (modes, all); return default_autovectorize_vector_modes (modes, all); @@ -10499,6 +10502,16 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) return false; } +/* Implements target hook vector_mode_supported_any_target_p. */ + +static bool +riscv_vector_mode_supported_any_target_p (machine_mode mode) +{ + if (TARGET_XTHEADVECTOR) + return false; + return true; +} + /* Initialize the GCC target structure. */ #undef TARGET_ASM_ALIGNED_HI_OP #define TARGET_ASM_ALIGNED_HI_OP "\t.half\t" @@ -10842,6 +10855,9 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) #undef TARGET_PREFERRED_ELSE_VALUE #define TARGET_PREFERRED_ELSE_VALUE riscv_preferred_else_value +#undef TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P +#define TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P riscv_vector_mode_supported_any_target_p + struct gcc_target targetm = TARGET_INITIALIZER; #include "gt-riscv.h" diff --git a/gcc/config/riscv/riscv_th_vector.h b/gcc/config/riscv/riscv_th_vector.h new file mode 100644 index 00000000000..6f47e0c90a4 --- /dev/null +++ b/gcc/config/riscv/riscv_th_vector.h @@ -0,0 +1,49 @@ +/* RISC-V 'XTheadVector' Extension intrinsics include file. + Copyright (C) 2022-2023 Free Software Foundation, Inc. + + This file is part of GCC. + + GCC is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published + by the Free Software Foundation; either version 3, or (at your + option) any later version. + + GCC is distributed in the hope that it will be useful, but WITHOUT + ANY WARRANTY; without even the implied warranty of MERCHANTABILITY + or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public + License for more details. + + Under Section 7 of GPL version 3, you are granted additional + permissions described in the GCC Runtime Library Exception, version + 3.1, as published by the Free Software Foundation. + + You should have received a copy of the GNU General Public License and + a copy of the GCC Runtime Library Exception along with this program; + see the files COPYING3 and COPYING.RUNTIME respectively. If not, see + <" target="_blank">" target="_blank">http://www.gnu.org/licenses/>. */ + +#ifndef __RISCV_TH_VECTOR_H +#define __RISCV_TH_VECTOR_H + +#include <stdint.h> +#include <stddef.h> + +#ifndef __riscv_xtheadvector +#error "XTheadVector intrinsics require the xtheadvector extension." +#else +#ifdef __cplusplus +extern "C" { +#endif + +/* NOTE: This implementation of riscv_th_vector.h is intentionally short. It does + not define the RVV types and intrinsic functions directly in C and C++ + code, but instead uses the following pragma to tell GCC to insert the + necessary type and function definitions itself. The net effect is the + same, and the file is a complete implementation of riscv_th_vector.h. */ +#pragma riscv intrinsic "vector" + +#ifdef __cplusplus +} +#endif // __cplusplus +#endif // __riscv_xtheadvector +#endif // __RISCV_TH_ECTOR_H diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md new file mode 100644 index 00000000000..f0aa196d93b --- /dev/null +++ b/gcc/config/riscv/thead-vector.md @@ -0,0 +1,142 @@ +(define_c_enum "unspec" [ + UNSPEC_TH_VWLDST +]) + +(define_mode_iterator V_VLS_VT [V VLS VT]) +(define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) + +(define_split + [(set (match_operand:V_VB_VLS_VT 0 "reg_or_mem_operand") + (match_operand:V_VB_VLS_VT 1 "reg_or_mem_operand"))] + "TARGET_XTHEADVECTOR" + [(const_int 0)] + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + }) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:V_VLS_VT 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:V_VLS_VT + [(match_operand:V_VLS_VT 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2")]) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:VB 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:VB + [(match_operand:VB 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2") + (set (attr "sew") (const_int 8)) + (set (attr "vlmul") (symbol_ref "riscv_vector::LMUL_1"))]) + +(define_insn "@th_vsetvl<mode>" + [(set (match_operand:P 0 "register_operand" "=r") + (unspec:P [(match_operand:P 1 "vector_length_operand" "rK") + (match_operand 2 "const_int_operand" "i") + (match_operand 3 "const_int_operand" "i")] UNSPEC_VSETVL)) + (set (reg:SI VL_REGNUM) + (unspec:SI [(match_dup 1) + (match_dup 2) + (match_dup 3)] UNSPEC_VSETVL)) + (set (reg:SI VTYPE_REGNUM) + (unspec:SI [(match_dup 2) + (match_dup 3)] UNSPEC_VSETVL))] + "TARGET_XTHEADVECTOR" + "vsetvli\t%0,%1,e%2,%m3" + [(set_attr "type" "vsetvl") + (set_attr "mode" "<MODE>") + (set (attr "sew") (symbol_ref "INTVAL (operands[2])")) + (set (attr "vlmul") (symbol_ref "INTVAL (operands[3])"))]) + +;; vsetvl zero,zero,vtype instruction. +;; This pattern has no side effects and does not set X0 register. +(define_insn "th_vsetvl_vtype_change_only" + [(set (reg:SI VTYPE_REGNUM) + (unspec:SI + [(match_operand 0 "const_int_operand" "i") + (match_operand 1 "const_int_operand" "i")] UNSPEC_VSETVL))] + "TARGET_XTHEADVECTOR" + "vsetvli\tzero,zero,e%0,%m1" + [(set_attr "type" "vsetvl") + (set_attr "mode" "SI") + (set (attr "sew") (symbol_ref "INTVAL (operands[0])")) + (set (attr "vlmul") (symbol_ref "INTVAL (operands[1])"))]) + +;; vsetvl zero,rs1,vtype instruction. +;; The reason we need this pattern since we should avoid setting X0 register +;; in vsetvl instruction pattern. +(define_insn "@th_vsetvl_discard_result<mode>" + [(set (reg:SI VL_REGNUM) + (unspec:SI [(match_operand:P 0 "vector_length_operand" "rK") + (match_operand 1 "const_int_operand" "i") + (match_operand 2 "const_int_operand" "i")] UNSPEC_VSETVL)) + (set (reg:SI VTYPE_REGNUM) + (unspec:SI [(match_dup 1) + (match_dup 2)] UNSPEC_VSETVL))] + "TARGET_XTHEADVECTOR" + "vsetvli\tzero,%0,e%1,%m2" + [(set_attr "type" "vsetvl") + (set_attr "mode" "<MODE>") + (set (attr "sew") (symbol_ref "INTVAL (operands[1])")) + (set (attr "vlmul") (symbol_ref "INTVAL (operands[2])"))]) + +;; It's emit by vsetvl/vsetvlmax intrinsics with no side effects. +;; Since we have many optmization passes from "expand" to "reload_completed", +;; such pattern can allow us gain benefits of these optimizations. +(define_insn_and_split "@th_vsetvl<mode>_no_side_effects" + [(set (match_operand:P 0 "register_operand" "=r") + (unspec:P [(match_operand:P 1 "vector_length_operand" "rK") + (match_operand 2 "const_int_operand" "i") + (match_operand 3 "const_int_operand" "i")] UNSPEC_VSETVL))] + "TARGET_XTHEADVECTOR" + "#" + "&& epilogue_completed" + [(parallel + [(set (match_dup 0) + (unspec:P [(match_dup 1) (match_dup 2) (match_dup 3)] UNSPEC_VSETVL)) + (set (reg:SI VL_REGNUM) + (unspec:SI [(match_dup 1) (match_dup 2) (match_dup 3)] UNSPEC_VSETVL)) + (set (reg:SI VTYPE_REGNUM) + (unspec:SI [(match_dup 2) (match_dup 3)] UNSPEC_VSETVL))])] + "" + [(set_attr "type" "vsetvl") + (set_attr "mode" "SI")]) diff --git a/gcc/config/riscv/vector-iterators.md b/gcc/config/riscv/vector-iterators.md index 5f5f7b5b986..c0fc7a2441d 100644 --- a/gcc/config/riscv/vector-iterators.md +++ b/gcc/config/riscv/vector-iterators.md @@ -109,11 +109,11 @@ ]) (define_mode_iterator VI [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -128,11 +128,11 @@ ;; allow the instruction and mode to be matched during combine et al. (define_mode_iterator VF [ (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -140,11 +140,11 @@ (define_mode_iterator VF_ZVFHMIN [ (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -271,16 +271,16 @@ ]) (define_mode_iterator VEEWEXT2 [ - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -290,10 +290,10 @@ ]) (define_mode_iterator VEEWEXT4 [ - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -311,59 +311,59 @@ ]) (define_mode_iterator VEEWTRUNC2 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") (RVVM4SI "TARGET_64BIT") (RVVM2SI "TARGET_64BIT") (RVVM1SI "TARGET_64BIT") - (RVVMF2SI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEEWTRUNC4 [ - RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM2HI "TARGET_64BIT") (RVVM1HI "TARGET_64BIT") - (RVVMF2HI "TARGET_64BIT") - (RVVMF4HI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2HI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") (RVVM1HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEEWTRUNC8 [ (RVVM1QI "TARGET_64BIT") - (RVVMF2QI "TARGET_64BIT") - (RVVMF4QI "TARGET_64BIT") - (RVVMF8QI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2QI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF4QI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEI16 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -452,11 +452,11 @@ ]) (define_mode_iterator VFULLI [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_FULL_V") (RVVM4DI "TARGET_FULL_V") (RVVM2DI "TARGET_FULL_V") (RVVM1DI "TARGET_FULL_V") @@ -509,17 +509,17 @@ ]) (define_mode_iterator VI_QH [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") ]) (define_mode_iterator VI_QHS [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -560,11 +560,11 @@ ]) (define_mode_iterator VI_QHS_NO_M8 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -603,11 +603,11 @@ (define_mode_iterator VF_HS [ (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (V1HF "riscv_vector::vls_mode_valid_p (V1HFmode) && TARGET_ZVFH") (V2HF "riscv_vector::vls_mode_valid_p (V2HFmode) && TARGET_ZVFH") @@ -638,12 +638,12 @@ (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") - (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (V1HF "riscv_vector::vls_mode_valid_p (V1HFmode) && TARGET_ZVFH") (V2HF "riscv_vector::vls_mode_valid_p (V2HFmode) && TARGET_ZVFH") @@ -674,11 +674,11 @@ ]) (define_mode_iterator V_VLSI_QHS [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -756,27 +756,27 @@ ;; E.g. when index mode = RVVM8QImde and Pmode = SImode, if it is not zero_extend or ;; scalar != 1, such gather/scatter is not allowed since we don't have RVVM32SImode. (define_mode_iterator RATIO64 [ - (RVVMF8QI "TARGET_MIN_VLEN > 32") - (RVVMF4HI "TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM1DI "TARGET_VECTOR_ELEN_64") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") ]) (define_mode_iterator RATIO32 [ - RVVMF4QI - RVVMF2HI + (RVVMF4QI "!TARGET_XTHEADVECTOR") + (RVVMF2HI "!TARGET_XTHEADVECTOR") RVVM1SI (RVVM2DI "TARGET_VECTOR_ELEN_64") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") ]) (define_mode_iterator RATIO16 [ - RVVMF2QI + (RVVMF2QI "!TARGET_XTHEADVECTOR") RVVM1HI RVVM2SI (RVVM4DI "TARGET_VECTOR_ELEN_64") @@ -814,21 +814,21 @@ ]) (define_mode_iterator RATIO64I [ - (RVVMF8QI "TARGET_MIN_VLEN > 32") - (RVVMF4HI "TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM1DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") ]) (define_mode_iterator RATIO32I [ - RVVMF4QI - RVVMF2HI + (RVVMF4QI "!TARGET_XTHEADVECTOR") + (RVVMF2HI "!TARGET_XTHEADVECTOR") RVVM1SI (RVVM2DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") ]) (define_mode_iterator RATIO16I [ - RVVMF2QI + (RVVMF2QI "!TARGET_XTHEADVECTOR") RVVM1HI RVVM2SI (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") @@ -873,21 +873,21 @@ ]) (define_mode_iterator V_FRACT [ - RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") ]) (define_mode_iterator VWEXTI [ - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -933,7 +933,7 @@ (RVVM4SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -966,7 +966,7 @@ (RVVM4SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -996,7 +996,7 @@ (define_mode_iterator VWCONVERTI [ (RVVM8SI "TARGET_ZVFH") (RVVM4SI "TARGET_ZVFH") (RVVM2SI "TARGET_ZVFH") (RVVM1SI "TARGET_ZVFH") - (RVVMF2SI "TARGET_ZVFH") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_ZVFH") (RVVM8DI "TARGET_VECTOR_ELEN_64 && TARGET_VECTOR_ELEN_FP_32") (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_VECTOR_ELEN_FP_32") @@ -1045,7 +1045,7 @@ ]) (define_mode_iterator VQEXTI [ - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -1456,11 +1456,11 @@ ;; VINDEXED [VI8 VI16 VI32 (VI64 "TARGET_64BIT")]. (define_mode_iterator VINDEXED [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") @@ -1468,12 +1468,12 @@ (RVVM1DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64 && TARGET_64BIT") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64 && TARGET_64BIT") @@ -3173,11 +3173,11 @@ (define_mode_iterator V_VLS_F_CONVERT_SI [ (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") - (RVVMF2HF "TARGET_ZVFH") (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") @@ -3290,12 +3290,12 @@ ]) (define_mode_iterator V_VLS_F_CONVERT_DI [ - (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index b5a9055cdc4..9e7a487759a 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -83,7 +83,7 @@ ;; check. However, we need default value of SEW for vsetvl instruction since there ;; is no field for ratio in the vsetvl instruction encoding. (define_attr "sew" "" - (cond [(eq_attr "mode" "RVVMF64BI,RVVMF32BI,RVVMF16BI,RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ + (cond [(eq_attr "mode" "RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ RVVM8QI,RVVM4QI,RVVM2QI,RVVM1QI,RVVMF2QI,RVVMF4QI,RVVMF8QI,\ RVVM1x8QI,RVVMF2x8QI,RVVMF4x8QI,RVVMF8x8QI,\ RVVM1x7QI,RVVMF2x7QI,RVVMF4x7QI,RVVMF8x7QI,\ @@ -95,6 +95,18 @@ V1QI,V2QI,V4QI,V8QI,V16QI,V32QI,V64QI,V128QI,V256QI,V512QI,V1024QI,V2048QI,V4096QI,\ V1BI,V2BI,V4BI,V8BI,V16BI,V32BI,V64BI,V128BI,V256BI,V512BI,V1024BI,V2048BI,V4096BI") (const_int 8) + (eq_attr "mode" "RVVMF16BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 16) + (const_int 8)) + (eq_attr "mode" "RVVMF32BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 32) + (const_int 8)) + (eq_attr "mode" "RVVMF64BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 64) + (const_int 8)) (eq_attr "mode" "RVVM8HI,RVVM4HI,RVVM2HI,RVVM1HI,RVVMF2HI,RVVMF4HI,\ RVVM1x8HI,RVVMF2x8HI,RVVMF4x8HI,\ RVVM1x7HI,RVVMF2x7HI,RVVMF4x7HI,\ @@ -155,9 +167,9 @@ (eq_attr "mode" "RVVM4QI,RVVMF2BI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2QI,RVVMF4BI") (symbol_ref "riscv_vector::LMUL_2") (eq_attr "mode" "RVVM1QI,RVVMF8BI") (symbol_ref "riscv_vector::LMUL_1") - (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "riscv_vector::LMUL_F2") - (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "riscv_vector::LMUL_F4") - (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "riscv_vector::LMUL_F8") + (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F2") + (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F4") + (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F8") (eq_attr "mode" "RVVM8HI") (symbol_ref "riscv_vector::LMUL_8") (eq_attr "mode" "RVVM4HI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2HI") (symbol_ref "riscv_vector::LMUL_2") @@ -428,6 +440,10 @@ vislide1up,vislide1down,vfslide1up,vfslide1down,\ vgather,vcompress,vlsegdux,vlsegdox,vssegtux,vssegtox") (const_int INVALID_ATTRIBUTE) + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE) (eq_attr "mode" "RVVM8QI,RVVM1BI") (const_int 1) (eq_attr "mode" "RVVM4QI,RVVMF2BI") (const_int 2) (eq_attr "mode" "RVVM2QI,RVVMF4BI") (const_int 4) @@ -888,6 +904,8 @@ (symbol_ref "riscv_vector::FRM_DYN")] (symbol_ref "riscv_vector::FRM_NONE"))) +(include "thead-vector.md") + ;; ----------------------------------------------------------------- ;; ---- Miscellaneous Operations ;; ----------------------------------------------------------------- @@ -1097,7 +1115,7 @@ (define_insn "*mov<mode>_whole" [(set (match_operand:V_WHOLE 0 "reg_or_mem_operand" "=vr, m,vr") (match_operand:V_WHOLE 1 "reg_or_mem_operand" " m,vr,vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "@ vl%m1re<sew>.v\t%0,%1 vs%m1r.v\t%1,%0 @@ -1125,7 +1143,7 @@ (define_insn "*mov<mode>" [(set (match_operand:VB 0 "register_operand" "=vr") (match_operand:VB 1 "register_operand" " vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "vmv1r.v\t%0,%1" [(set_attr "type" "vmov") (set_attr "mode" "<MODE>")]) @@ -3680,7 +3698,7 @@ (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3701,7 +3719,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3722,7 +3740,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c index 2e0e12aa045..2eef9e1e1a8 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c @@ -1,4 +1,4 @@ -/* { dg-do compile } */ +/* { dg-do compile { target { ! riscv_xtheadvector } } } */ /* { dg-skip-if "test rvv intrinsic" { *-*-* } { "*" } { "-march=rv*v*" } } */ void foo0 () {__rvv_bool64_t t;} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c index 3d81b179235..ef329e30785 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c @@ -1,4 +1,4 @@ /* { dg-do compile } */ /* { dg-options "-O3 -march=rv32gc -mabi=ilp32d" } */ -#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' extension enabled} } */ +#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' or 'XTHEADVECTOR' extension enabled} } */ diff --git a/gcc/testsuite/lib/target-supports.exp b/gcc/testsuite/lib/target-supports.exp index 7f13ff0ca56..70df6b1401c 100644 --- a/gcc/testsuite/lib/target-supports.exp +++ b/gcc/testsuite/lib/target-supports.exp @@ -1952,6 +1952,18 @@ proc check_effective_target_riscv_zbb { } { }] } +# Return 1 if the target arch supports the XTheadVector extension, 0 otherwise. +# Cache the result. + +proc check_effective_target_riscv_xtheadvector { } { + return [check_no_compiler_messages riscv_ext_xtheadvector assembly { + #ifndef __riscv_xtheadvector + #error "Not __riscv_xtheadvector" + #endif + }] +} + + # Return 1 if we can execute code when using dg-add-options riscv_v proc check_effective_target_riscv_v_ok { } { -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: Re：[PATCH v4] RISC-V: Handle differences between XTheadvector and Vector 2024-01-02 3:40 ` Re：Re：[PATCH " joshua @ 2024-01-02 3:41 ` juzhe.zhong 0 siblings, 0 replies; 133+ messages in thread From: juzhe.zhong @ 2024-01-02 3:41 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner [-- Attachment #1: Type: text/plain, Size: 73524 bytes --] OK. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-02 11:40 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner 主题： Re：Re：[PATCH v4] RISC-V: Handle differences between XTheadvector and Vector But the riscv_print_operand() function returns void. We cannot return instructions like riscv_output_move. I think the briefest approach is to add some logic in the vsetvl patterns. "TARGET_VECTOR" { return TARGET_XTHEADVECTOR ? "vsetvli\t%0,%1,e%2,%m3" : "vset%i1vli\t%0,%1,e%2,%m3,t%p4,m%p5"; } Only 3 patterns need to be modified and I don't think it is too invasive. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月2日(星期二) 11:10 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu> 主 题：Re: Re：[PATCH v4] RISC-V: Handle differences between XTheadvector and Vector Like riscv_output_move if (TARGET_THEADVECTOR) return vsetvli....no tail policy and mask policy. else return .... juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-02 11:03 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner 主题： Re：[PATCH v4] RISC-V: Handle differences between XTheadvector and Vector For vsetvl issues, what we want to do here is to directly remove "t" and "m". If we add TARGET_XTHEADVECTOR logic in case "p" in riscv_print_operand, how can we remove "t" and "m"? If I use "break", assembly like "th.vsetvli zero,a5,e8,m1,t,m" will be returned. if (TARGET_THEADVECTOR) ... else else if (code == CONST_INT) { /* Tail && Mask policy. */ asm_fprintf (file, "%s", IS_AGNOSTIC (UINTVAL (op)) ? "a" : "u"); } ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月2日(星期二) 10:00 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v4] RISC-V: Handle differences between XTheadvector and Vector + if (TARGET_XTHEADVECTOR) + return false; Move it to : if (TARGET_VECTOR && stringop_strategy & STRATEGY_VECTOR) { bool ok = riscv_vector::expand_block_move (dest, src, length); if (ok) return true; } (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (match_test "!TARGET_XTHEADVECTOR || rtx_equal_p (op, const0_rtx)") + (match_operand 0 "const_csr_operand")))) It's hard to trace. Change it into : (ior 1. TARGET_THEADVECTOR && rtx_equal_p (op, const0_rtx) 2. !TAGEET_THEADVECTOR && const_csr_operand) + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (mode, dest, src, + RVV_VLMAX, GEN_INT(VLMAX))); + return true; + } Move it outside legitimize_move It should be it: if (TARGET_THEADVECTOR) emit_th_move... DONE; else if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; vsetvli issues: I wonder whether we can use ASM_OUTPUT_OPCODE to recognize "ta,ma"/"ta,mu"/"tu,ma"/"tu,mu" and replace these 4 variants by "". So that we don't have tail policy and mask policy in vsetvli ASM string. Another alternative approach is we can change vsetlvi ASM rule: "vset%i1vli\t%0,%1,e%2,%m3,t%p4,m%p5" if (TARGET_THEADVECTOR) ... else else if (code == CONST_INT) { /* Tail && Mask policy. */ asm_fprintf (file, "%s", IS_AGNOSTIC (UINTVAL (op)) ? "a" : "u"); } in riscv.cc. The benefit is that we can avoid adding all th_vsetvl patterns and invasive code changs in VSETVL PASS. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2023-12-29 12:21 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v4] RISC-V: Handle differences between XTheadvector and Vector This patch is to handle the differences in instruction generation between Vector and XTheadVector. In this version, we only support partial xtheadvector instructions that leverage directly from current RVV1.0 with simple adding "th." prefix. For different name xtheadvector instructions but share same patterns as RVV1.0 instructions, we will use ASM targethook to rewrite the whole string of the instructions in the following patches. For some vector patterns that cannot be avoided, we use "!TARGET_XTHEADVECTOR" to disable them in vector.md in order not to generate instructions that xtheadvector does not support, like vmv1r and vsext.vf2. gcc/ChangeLog: * config.gcc: Add files for XTheadVector intrinsics. * config/riscv/autovec.md: Guard XTheadVector. * config/riscv/riscv-string.cc (expand_block_move): Guard XTheadVector. * config/riscv/riscv-v.cc (legitimize_move): New expansion. (get_prefer_tail_policy): Give specific value for tail. (get_prefer_mask_policy): Give specific value for mask. (vls_mode_valid_p): Avoid autovec. * config/riscv/riscv-vector-builtins-shapes.cc (check_type): (build_one): New function. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): (DEF_THEAD_RVV_FUNCTION): Add new marcos. (check_required_extensions): (handle_pragma_vector): * config/riscv/riscv-vector-builtins.h (RVV_REQUIRE_VECTOR): (RVV_REQUIRE_XTHEADVECTOR): Add RVV_REQUIRE_VECTOR and RVV_REQUIRE_XTHEADVECTOR. (struct function_group_info): * config/riscv/riscv-vector-switch.def (ENTRY): Disable fractional mode for the XTheadVector extension. (TUPLE_ENTRY): Likewise. * config/riscv/riscv-vsetvl.cc: Add functions for xtheadvector. * config/riscv/riscv.cc (riscv_v_ext_vls_mode_p): Guard XTheadVector. (riscv_v_adjust_bytesize): Likewise. (riscv_preferred_simd_mode): Likewsie. (riscv_autovectorize_vector_modes): Likewise. (riscv_vector_mode_supported_any_target_p): Likewise. (TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P): Likewise. * config/riscv/vector-iterators.md: Remove fractional LMUL. * config/riscv/vector.md: Include thead-vector.md. * config/riscv/riscv_th_vector.h: New file. * config/riscv/thead-vector.md: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/base/pragma-1.c: Add XTheadVector. * gcc.target/riscv/rvv/base/abi-1.c: Exclude XTheadVector. * lib/target-supports.exp: Add target for XTheadVector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config.gcc | 2 +- gcc/config/riscv/autovec.md | 2 +- gcc/config/riscv/predicates.md | 3 +- gcc/config/riscv/riscv-string.cc | 3 + gcc/config/riscv/riscv-v.cc | 13 +- .../riscv/riscv-vector-builtins-bases.cc | 3 + .../riscv/riscv-vector-builtins-shapes.cc | 23 +++ gcc/config/riscv/riscv-vector-switch.def | 150 +++++++------- gcc/config/riscv/riscv-vsetvl.cc | 10 + gcc/config/riscv/riscv.cc | 20 +- gcc/config/riscv/riscv_th_vector.h | 49 +++++ gcc/config/riscv/thead-vector.md | 142 +++++++++++++ gcc/config/riscv/vector-iterators.md | 186 +++++++++--------- gcc/config/riscv/vector.md | 36 +++- .../gcc.target/riscv/rvv/base/abi-1.c | 2 +- .../gcc.target/riscv/rvv/base/pragma-1.c | 2 +- gcc/testsuite/lib/target-supports.exp | 12 ++ 17 files changed, 471 insertions(+), 187 deletions(-) create mode 100644 gcc/config/riscv/riscv_th_vector.h create mode 100644 gcc/config/riscv/thead-vector.md diff --git a/gcc/config.gcc b/gcc/config.gcc index f0676c830e8..1445d98c147 100644 --- a/gcc/config.gcc +++ b/gcc/config.gcc @@ -549,7 +549,7 @@ riscv*) extra_objs="${extra_objs} riscv-vector-builtins.o riscv-vector-builtins-shapes.o riscv-vector-builtins-bases.o" extra_objs="${extra_objs} thead.o riscv-target-attr.o" d_target_objs="riscv-d.o" - extra_headers="riscv_vector.h" + extra_headers="riscv_vector.h riscv_th_vector.h" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.cc" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.h" ;; diff --git a/gcc/config/riscv/autovec.md b/gcc/config/riscv/autovec.md index 8b8a92f10a1..1fac56c7095 100644 --- a/gcc/config/riscv/autovec.md +++ b/gcc/config/riscv/autovec.md @@ -2579,7 +2579,7 @@ [(match_operand 0 "register_operand") (match_operand 1 "memory_operand") (match_operand:ANYI 2 "const_int_operand")] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" { riscv_vector::expand_rawmemchr(<MODE>mode, operands[0], operands[1], operands[2]); diff --git a/gcc/config/riscv/predicates.md b/gcc/config/riscv/predicates.md index b87a6900841..c5ab86d5db9 100644 --- a/gcc/config/riscv/predicates.md +++ b/gcc/config/riscv/predicates.md @@ -428,7 +428,8 @@ ;; Predicates for the V extension. (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (match_test "!TARGET_XTHEADVECTOR || rtx_equal_p (op, const0_rtx)") + (match_operand 0 "const_csr_operand")))) (define_special_predicate "autovec_length_operand" (ior (match_operand 0 "pmode_register_operand") diff --git a/gcc/config/riscv/riscv-string.cc b/gcc/config/riscv/riscv-string.cc index 11c1f74d0b3..ec8f3486fd8 100644 --- a/gcc/config/riscv/riscv-string.cc +++ b/gcc/config/riscv/riscv-string.cc @@ -808,6 +808,9 @@ expand_block_move (rtx dst_in, rtx src_in, rtx length_in) bnez a2, loop # Any more? ret # Return */ + if (TARGET_XTHEADVECTOR) + return false; + gcc_assert (TARGET_VECTOR); HOST_WIDE_INT potential_ew diff --git a/gcc/config/riscv/riscv-v.cc b/gcc/config/riscv/riscv-v.cc index 038ab084a37..5e9e45aecd2 100644 --- a/gcc/config/riscv/riscv-v.cc +++ b/gcc/config/riscv/riscv-v.cc @@ -1523,6 +1523,13 @@ legitimize_move (rtx dest, rtx *srcp) return true; } + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (mode, dest, src, + RVV_VLMAX, GEN_INT(VLMAX))); + return true; + } + if (riscv_v_ext_vls_mode_p (mode)) { if (GET_MODE_NUNITS (mode).to_constant () <= 31) @@ -1772,7 +1779,7 @@ get_prefer_tail_policy () compiler pick up either agnostic or undisturbed. Maybe we will have a compile option like -mprefer=agnostic to set this value???. */ - return TAIL_ANY; + return TARGET_XTHEADVECTOR ? TAIL_AGNOSTIC : TAIL_ANY; } /* Get prefer mask policy. */ @@ -1783,7 +1790,7 @@ get_prefer_mask_policy () compiler pick up either agnostic or undisturbed. Maybe we will have a compile option like -mprefer=agnostic to set this value???. */ - return MASK_ANY; + return TARGET_XTHEADVECTOR ? MASK_UNDISTURBED : MASK_ANY; } /* Get avl_type rtx. */ @@ -4383,7 +4390,7 @@ cmp_lmul_gt_one (machine_mode mode) bool vls_mode_valid_p (machine_mode vls_mode) { - if (!TARGET_VECTOR) + if (!TARGET_VECTOR || TARGET_XTHEADVECTOR) return false; if (riscv_autovec_preference == RVV_SCALABLE) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index c51affde353..2918c07ebf3 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -133,6 +133,9 @@ public: = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + if (TARGET_XTHEADVECTOR) + return e.generate_insn (code_for_th_vsetvl_no_side_effects (Pmode)); + /* TAIL_ANY. */ e.add_input_operand (Pmode, gen_int_mode (get_prefer_tail_policy (), Pmode)); diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 4a754e0228f..6b49404a1fa 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -33,6 +33,25 @@ namespace riscv_vector { +/* Check whether the RETURN_TYPE and ARGUMENT_TYPES are + valid for the function. */ + +static bool +check_type (tree return_type, vec<tree> &argument_types) +{ + tree arg; + unsigned i; + + if (!return_type) + return false; + + FOR_EACH_VEC_ELT (argument_types, i, arg) + if (!arg) + return false; + + return true; +} + /* Add one function instance for GROUP, using operand suffix at index OI, mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ static void @@ -49,6 +68,10 @@ build_one (function_builder &b, const function_group_info &group, group.ops_infos.types[vec_type_idx].index); b.allocate_argument_types (function_instance, argument_types); b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + b.add_overloaded_function (function_instance, *group.shape); b.add_unique_function (function_instance, (*group.shape), return_type, argument_types); diff --git a/gcc/config/riscv/riscv-vector-switch.def b/gcc/config/riscv/riscv-vector-switch.def index 5c9f9bcbc3e..f7a66b34bae 100644 --- a/gcc/config/riscv/riscv-vector-switch.def +++ b/gcc/config/riscv/riscv-vector-switch.def @@ -68,9 +68,9 @@ Encode the ratio of SEW/LMUL into the mask types. #endif /* Disable modes if TARGET_MIN_VLEN == 32. */ -ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) -ENTRY (RVVMF32BI, true, LMUL_F4, 32) -ENTRY (RVVMF16BI, true, LMUL_F2, 16) +ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F8, 64) +ENTRY (RVVMF32BI, true, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F4, 32) +ENTRY (RVVMF16BI, true, TARGET_XTHEADVECTOR ? LMUL_1 : LMUL_F2 , 16) ENTRY (RVVMF8BI, true, LMUL_1, 8) ENTRY (RVVMF4BI, true, LMUL_2, 4) ENTRY (RVVMF2BI, true, LMUL_4, 2) @@ -81,39 +81,39 @@ ENTRY (RVVM8QI, true, LMUL_8, 1) ENTRY (RVVM4QI, true, LMUL_4, 2) ENTRY (RVVM2QI, true, LMUL_2, 4) ENTRY (RVVM1QI, true, LMUL_1, 8) -ENTRY (RVVMF2QI, true, LMUL_F2, 16) -ENTRY (RVVMF4QI, true, LMUL_F4, 32) -ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) +ENTRY (RVVMF2QI, !TARGET_XTHEADVECTOR, LMUL_F2, 16) +ENTRY (RVVMF4QI, !TARGET_XTHEADVECTOR, LMUL_F4, 32) +ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F8, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8HI, true, LMUL_8, 2) ENTRY (RVVM4HI, true, LMUL_4, 4) ENTRY (RVVM2HI, true, LMUL_2, 8) ENTRY (RVVM1HI, true, LMUL_1, 16) -ENTRY (RVVMF2HI, true, LMUL_F2, 32) -ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HI, !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_16. */ ENTRY (RVVM8HF, TARGET_VECTOR_ELEN_FP_16, LMUL_8, 2) ENTRY (RVVM4HF, TARGET_VECTOR_ELEN_FP_16, LMUL_4, 4) ENTRY (RVVM2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_2, 8) ENTRY (RVVM1HF, TARGET_VECTOR_ELEN_FP_16, LMUL_1, 16) -ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_F2, 32) -ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8SI, true, LMUL_8, 4) ENTRY (RVVM4SI, true, LMUL_4, 8) ENTRY (RVVM2SI, true, LMUL_2, 16) ENTRY (RVVM1SI, true, LMUL_1, 32) -ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_32. */ ENTRY (RVVM8SF, TARGET_VECTOR_ELEN_FP_32, LMUL_8, 4) ENTRY (RVVM4SF, TARGET_VECTOR_ELEN_FP_32, LMUL_4, 8) ENTRY (RVVM2SF, TARGET_VECTOR_ELEN_FP_32, LMUL_2, 16) ENTRY (RVVM1SF, TARGET_VECTOR_ELEN_FP_32, LMUL_1, 32) -ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if !TARGET_VECTOR_ELEN_64. */ ENTRY (RVVM8DI, TARGET_VECTOR_ELEN_64, LMUL_8, 8) @@ -140,127 +140,127 @@ ENTRY (RVVM1DF, TARGET_VECTOR_ELEN_FP_64, LMUL_1, 64) #endif TUPLE_ENTRY (RVVM1x8QI, true, RVVM1QI, 8, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x8QI, true, RVVMF2QI, 8, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x8QI, true, RVVMF4QI, 8, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 8, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x8QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 8, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x8QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 8, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 8, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x7QI, true, RVVM1QI, 7, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x7QI, true, RVVMF2QI, 7, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x7QI, true, RVVMF4QI, 7, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 7, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x7QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 7, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x7QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 7, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 7, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x6QI, true, RVVM1QI, 6, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x6QI, true, RVVMF2QI, 6, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x6QI, true, RVVMF4QI, 6, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 6, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x6QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 6, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x6QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 6, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 6, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x5QI, true, RVVM1QI, 5, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x5QI, true, RVVMF2QI, 5, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x5QI, true, RVVMF4QI, 5, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 5, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x5QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 5, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x5QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 5, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 5, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x4QI, true, RVVM2QI, 4, LMUL_2, 4) TUPLE_ENTRY (RVVM1x4QI, true, RVVM1QI, 4, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x4QI, true, RVVMF2QI, 4, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x4QI, true, RVVMF4QI, 4, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 4, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x4QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 4, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x4QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 4, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 4, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x3QI, true, RVVM2QI, 3, LMUL_2, 4) TUPLE_ENTRY (RVVM1x3QI, true, RVVM1QI, 3, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x3QI, true, RVVMF2QI, 3, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x3QI, true, RVVMF4QI, 3, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 3, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x3QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 3, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x3QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 3, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 3, LMUL_F8, 64) TUPLE_ENTRY (RVVM4x2QI, true, RVVM4QI, 2, LMUL_4, 2) TUPLE_ENTRY (RVVM2x2QI, true, RVVM2QI, 2, LMUL_2, 4) TUPLE_ENTRY (RVVM1x2QI, true, RVVM1QI, 2, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x2QI, true, RVVMF2QI, 2, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x2QI, true, RVVMF4QI, 2, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 2, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x2QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 2, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x2QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 2, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 2, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x8HI, true, RVVM1HI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HI, true, RVVMF2HI, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HI, true, RVVM1HI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HI, true, RVVMF2HI, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HI, true, RVVM1HI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HI, true, RVVMF2HI, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HI, true, RVVM1HI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HI, true, RVVMF2HI, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HI, true, RVVM2HI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HI, true, RVVM1HI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HI, true, RVVMF2HI, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HI, true, RVVM2HI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HI, true, RVVM1HI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HI, true, RVVMF2HI, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HI, true, RVVM4HI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HI, true, RVVM2HI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HI, true, RVVM1HI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HI, true, RVVMF2HI, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM4HF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8SI, true, RVVM1SI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SI, true, RVVM1SI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SI, true, RVVM1SI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SI, true, RVVM1SI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SI, true, RVVM2SI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SI, true, RVVM1SI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SI, true, RVVM2SI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SI, true, RVVM1SI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SI, true, RVVM4SI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SI, true, RVVM2SI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SI, true, RVVM1SI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM4SF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 8, LMUL_1, 16) TUPLE_ENTRY (RVVM1x7DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 7, LMUL_1, 16) diff --git a/gcc/config/riscv/riscv-vsetvl.cc b/gcc/config/riscv/riscv-vsetvl.cc index eabaef80f89..c726253c107 100644 --- a/gcc/config/riscv/riscv-vsetvl.cc +++ b/gcc/config/riscv/riscv-vsetvl.cc @@ -1117,6 +1117,16 @@ public: avl = GEN_INT (0); rtx sew = gen_int_mode (get_sew (), Pmode); rtx vlmul = gen_int_mode (get_vlmul (), Pmode); + + if (TARGET_XTHEADVECTOR) { + if (change_vtype_only_p ()) + return gen_th_vsetvl_vtype_change_only (sew, vlmul); + else if (has_vl () && !ignore_vl) + return gen_th_vsetvl (Pmode, get_vl (), avl, sew, vlmul); + else + return gen_th_vsetvl_discard_result (Pmode, avl, sew, vlmul); + } + rtx ta = gen_int_mode (get_ta (), Pmode); rtx ma = gen_int_mode (get_ma (), Pmode); diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index ea1d59d9cf2..a80bf8d1a74 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -1406,6 +1406,9 @@ riscv_v_adjust_bytesize (machine_mode mode, int scale) { if (riscv_v_ext_vector_mode_p (mode)) { + if (TARGET_XTHEADVECTOR) + return BYTES_PER_RISCV_VECTOR; + poly_int64 nunits = GET_MODE_NUNITS (mode); poly_int64 mode_size = GET_MODE_SIZE (mode); @@ -9965,7 +9968,7 @@ riscv_use_divmod_expander (void) static machine_mode riscv_preferred_simd_mode (scalar_mode mode) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::preferred_simd_mode (mode); return word_mode; @@ -10316,7 +10319,7 @@ riscv_mode_priority (int, int n) unsigned int riscv_autovectorize_vector_modes (vector_modes *modes, bool all) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::autovectorize_vector_modes (modes, all); return default_autovectorize_vector_modes (modes, all); @@ -10499,6 +10502,16 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) return false; } +/* Implements target hook vector_mode_supported_any_target_p. */ + +static bool +riscv_vector_mode_supported_any_target_p (machine_mode mode) +{ + if (TARGET_XTHEADVECTOR) + return false; + return true; +} + /* Initialize the GCC target structure. */ #undef TARGET_ASM_ALIGNED_HI_OP #define TARGET_ASM_ALIGNED_HI_OP "\t.half\t" @@ -10842,6 +10855,9 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) #undef TARGET_PREFERRED_ELSE_VALUE #define TARGET_PREFERRED_ELSE_VALUE riscv_preferred_else_value +#undef TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P +#define TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P riscv_vector_mode_supported_any_target_p + struct gcc_target targetm = TARGET_INITIALIZER; #include "gt-riscv.h" diff --git a/gcc/config/riscv/riscv_th_vector.h b/gcc/config/riscv/riscv_th_vector.h new file mode 100644 index 00000000000..6f47e0c90a4 --- /dev/null +++ b/gcc/config/riscv/riscv_th_vector.h @@ -0,0 +1,49 @@ +/* RISC-V 'XTheadVector' Extension intrinsics include file. + Copyright (C) 2022-2023 Free Software Foundation, Inc. + + This file is part of GCC. + + GCC is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published + by the Free Software Foundation; either version 3, or (at your + option) any later version. + + GCC is distributed in the hope that it will be useful, but WITHOUT + ANY WARRANTY; without even the implied warranty of MERCHANTABILITY + or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public + License for more details. + + Under Section 7 of GPL version 3, you are granted additional + permissions described in the GCC Runtime Library Exception, version + 3.1, as published by the Free Software Foundation. + + You should have received a copy of the GNU General Public License and + a copy of the GCC Runtime Library Exception along with this program; + see the files COPYING3 and COPYING.RUNTIME respectively. If not, see + <" target="_blank">" target="_blank">http://www.gnu.org/licenses/>. */ + +#ifndef __RISCV_TH_VECTOR_H +#define __RISCV_TH_VECTOR_H + +#include <stdint.h> +#include <stddef.h> + +#ifndef __riscv_xtheadvector +#error "XTheadVector intrinsics require the xtheadvector extension." +#else +#ifdef __cplusplus +extern "C" { +#endif + +/* NOTE: This implementation of riscv_th_vector.h is intentionally short. It does + not define the RVV types and intrinsic functions directly in C and C++ + code, but instead uses the following pragma to tell GCC to insert the + necessary type and function definitions itself. The net effect is the + same, and the file is a complete implementation of riscv_th_vector.h. */ +#pragma riscv intrinsic "vector" + +#ifdef __cplusplus +} +#endif // __cplusplus +#endif // __riscv_xtheadvector +#endif // __RISCV_TH_ECTOR_H diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md new file mode 100644 index 00000000000..f0aa196d93b --- /dev/null +++ b/gcc/config/riscv/thead-vector.md @@ -0,0 +1,142 @@ +(define_c_enum "unspec" [ + UNSPEC_TH_VWLDST +]) + +(define_mode_iterator V_VLS_VT [V VLS VT]) +(define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) + +(define_split + [(set (match_operand:V_VB_VLS_VT 0 "reg_or_mem_operand") + (match_operand:V_VB_VLS_VT 1 "reg_or_mem_operand"))] + "TARGET_XTHEADVECTOR" + [(const_int 0)] + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + }) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:V_VLS_VT 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:V_VLS_VT + [(match_operand:V_VLS_VT 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2")]) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:VB 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:VB + [(match_operand:VB 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2") + (set (attr "sew") (const_int 8)) + (set (attr "vlmul") (symbol_ref "riscv_vector::LMUL_1"))]) + +(define_insn "@th_vsetvl<mode>" + [(set (match_operand:P 0 "register_operand" "=r") + (unspec:P [(match_operand:P 1 "vector_length_operand" "rK") + (match_operand 2 "const_int_operand" "i") + (match_operand 3 "const_int_operand" "i")] UNSPEC_VSETVL)) + (set (reg:SI VL_REGNUM) + (unspec:SI [(match_dup 1) + (match_dup 2) + (match_dup 3)] UNSPEC_VSETVL)) + (set (reg:SI VTYPE_REGNUM) + (unspec:SI [(match_dup 2) + (match_dup 3)] UNSPEC_VSETVL))] + "TARGET_XTHEADVECTOR" + "vsetvli\t%0,%1,e%2,%m3" + [(set_attr "type" "vsetvl") + (set_attr "mode" "<MODE>") + (set (attr "sew") (symbol_ref "INTVAL (operands[2])")) + (set (attr "vlmul") (symbol_ref "INTVAL (operands[3])"))]) + +;; vsetvl zero,zero,vtype instruction. +;; This pattern has no side effects and does not set X0 register. +(define_insn "th_vsetvl_vtype_change_only" + [(set (reg:SI VTYPE_REGNUM) + (unspec:SI + [(match_operand 0 "const_int_operand" "i") + (match_operand 1 "const_int_operand" "i")] UNSPEC_VSETVL))] + "TARGET_XTHEADVECTOR" + "vsetvli\tzero,zero,e%0,%m1" + [(set_attr "type" "vsetvl") + (set_attr "mode" "SI") + (set (attr "sew") (symbol_ref "INTVAL (operands[0])")) + (set (attr "vlmul") (symbol_ref "INTVAL (operands[1])"))]) + +;; vsetvl zero,rs1,vtype instruction. +;; The reason we need this pattern since we should avoid setting X0 register +;; in vsetvl instruction pattern. +(define_insn "@th_vsetvl_discard_result<mode>" + [(set (reg:SI VL_REGNUM) + (unspec:SI [(match_operand:P 0 "vector_length_operand" "rK") + (match_operand 1 "const_int_operand" "i") + (match_operand 2 "const_int_operand" "i")] UNSPEC_VSETVL)) + (set (reg:SI VTYPE_REGNUM) + (unspec:SI [(match_dup 1) + (match_dup 2)] UNSPEC_VSETVL))] + "TARGET_XTHEADVECTOR" + "vsetvli\tzero,%0,e%1,%m2" + [(set_attr "type" "vsetvl") + (set_attr "mode" "<MODE>") + (set (attr "sew") (symbol_ref "INTVAL (operands[1])")) + (set (attr "vlmul") (symbol_ref "INTVAL (operands[2])"))]) + +;; It's emit by vsetvl/vsetvlmax intrinsics with no side effects. +;; Since we have many optmization passes from "expand" to "reload_completed", +;; such pattern can allow us gain benefits of these optimizations. +(define_insn_and_split "@th_vsetvl<mode>_no_side_effects" + [(set (match_operand:P 0 "register_operand" "=r") + (unspec:P [(match_operand:P 1 "vector_length_operand" "rK") + (match_operand 2 "const_int_operand" "i") + (match_operand 3 "const_int_operand" "i")] UNSPEC_VSETVL))] + "TARGET_XTHEADVECTOR" + "#" + "&& epilogue_completed" + [(parallel + [(set (match_dup 0) + (unspec:P [(match_dup 1) (match_dup 2) (match_dup 3)] UNSPEC_VSETVL)) + (set (reg:SI VL_REGNUM) + (unspec:SI [(match_dup 1) (match_dup 2) (match_dup 3)] UNSPEC_VSETVL)) + (set (reg:SI VTYPE_REGNUM) + (unspec:SI [(match_dup 2) (match_dup 3)] UNSPEC_VSETVL))])] + "" + [(set_attr "type" "vsetvl") + (set_attr "mode" "SI")]) diff --git a/gcc/config/riscv/vector-iterators.md b/gcc/config/riscv/vector-iterators.md index 5f5f7b5b986..c0fc7a2441d 100644 --- a/gcc/config/riscv/vector-iterators.md +++ b/gcc/config/riscv/vector-iterators.md @@ -109,11 +109,11 @@ ]) (define_mode_iterator VI [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -128,11 +128,11 @@ ;; allow the instruction and mode to be matched during combine et al. (define_mode_iterator VF [ (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -140,11 +140,11 @@ (define_mode_iterator VF_ZVFHMIN [ (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -271,16 +271,16 @@ ]) (define_mode_iterator VEEWEXT2 [ - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -290,10 +290,10 @@ ]) (define_mode_iterator VEEWEXT4 [ - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -311,59 +311,59 @@ ]) (define_mode_iterator VEEWTRUNC2 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") (RVVM4SI "TARGET_64BIT") (RVVM2SI "TARGET_64BIT") (RVVM1SI "TARGET_64BIT") - (RVVMF2SI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEEWTRUNC4 [ - RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM2HI "TARGET_64BIT") (RVVM1HI "TARGET_64BIT") - (RVVMF2HI "TARGET_64BIT") - (RVVMF4HI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2HI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") (RVVM1HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEEWTRUNC8 [ (RVVM1QI "TARGET_64BIT") - (RVVMF2QI "TARGET_64BIT") - (RVVMF4QI "TARGET_64BIT") - (RVVMF8QI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2QI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF4QI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEI16 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -452,11 +452,11 @@ ]) (define_mode_iterator VFULLI [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_FULL_V") (RVVM4DI "TARGET_FULL_V") (RVVM2DI "TARGET_FULL_V") (RVVM1DI "TARGET_FULL_V") @@ -509,17 +509,17 @@ ]) (define_mode_iterator VI_QH [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") ]) (define_mode_iterator VI_QHS [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -560,11 +560,11 @@ ]) (define_mode_iterator VI_QHS_NO_M8 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -603,11 +603,11 @@ (define_mode_iterator VF_HS [ (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (V1HF "riscv_vector::vls_mode_valid_p (V1HFmode) && TARGET_ZVFH") (V2HF "riscv_vector::vls_mode_valid_p (V2HFmode) && TARGET_ZVFH") @@ -638,12 +638,12 @@ (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") - (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (V1HF "riscv_vector::vls_mode_valid_p (V1HFmode) && TARGET_ZVFH") (V2HF "riscv_vector::vls_mode_valid_p (V2HFmode) && TARGET_ZVFH") @@ -674,11 +674,11 @@ ]) (define_mode_iterator V_VLSI_QHS [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -756,27 +756,27 @@ ;; E.g. when index mode = RVVM8QImde and Pmode = SImode, if it is not zero_extend or ;; scalar != 1, such gather/scatter is not allowed since we don't have RVVM32SImode. (define_mode_iterator RATIO64 [ - (RVVMF8QI "TARGET_MIN_VLEN > 32") - (RVVMF4HI "TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM1DI "TARGET_VECTOR_ELEN_64") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") ]) (define_mode_iterator RATIO32 [ - RVVMF4QI - RVVMF2HI + (RVVMF4QI "!TARGET_XTHEADVECTOR") + (RVVMF2HI "!TARGET_XTHEADVECTOR") RVVM1SI (RVVM2DI "TARGET_VECTOR_ELEN_64") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") ]) (define_mode_iterator RATIO16 [ - RVVMF2QI + (RVVMF2QI "!TARGET_XTHEADVECTOR") RVVM1HI RVVM2SI (RVVM4DI "TARGET_VECTOR_ELEN_64") @@ -814,21 +814,21 @@ ]) (define_mode_iterator RATIO64I [ - (RVVMF8QI "TARGET_MIN_VLEN > 32") - (RVVMF4HI "TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM1DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") ]) (define_mode_iterator RATIO32I [ - RVVMF4QI - RVVMF2HI + (RVVMF4QI "!TARGET_XTHEADVECTOR") + (RVVMF2HI "!TARGET_XTHEADVECTOR") RVVM1SI (RVVM2DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") ]) (define_mode_iterator RATIO16I [ - RVVMF2QI + (RVVMF2QI "!TARGET_XTHEADVECTOR") RVVM1HI RVVM2SI (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") @@ -873,21 +873,21 @@ ]) (define_mode_iterator V_FRACT [ - RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") ]) (define_mode_iterator VWEXTI [ - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -933,7 +933,7 @@ (RVVM4SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -966,7 +966,7 @@ (RVVM4SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -996,7 +996,7 @@ (define_mode_iterator VWCONVERTI [ (RVVM8SI "TARGET_ZVFH") (RVVM4SI "TARGET_ZVFH") (RVVM2SI "TARGET_ZVFH") (RVVM1SI "TARGET_ZVFH") - (RVVMF2SI "TARGET_ZVFH") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_ZVFH") (RVVM8DI "TARGET_VECTOR_ELEN_64 && TARGET_VECTOR_ELEN_FP_32") (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_VECTOR_ELEN_FP_32") @@ -1045,7 +1045,7 @@ ]) (define_mode_iterator VQEXTI [ - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -1456,11 +1456,11 @@ ;; VINDEXED [VI8 VI16 VI32 (VI64 "TARGET_64BIT")]. (define_mode_iterator VINDEXED [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") @@ -1468,12 +1468,12 @@ (RVVM1DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64 && TARGET_64BIT") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64 && TARGET_64BIT") @@ -3173,11 +3173,11 @@ (define_mode_iterator V_VLS_F_CONVERT_SI [ (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") - (RVVMF2HF "TARGET_ZVFH") (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") @@ -3290,12 +3290,12 @@ ]) (define_mode_iterator V_VLS_F_CONVERT_DI [ - (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index b5a9055cdc4..9e7a487759a 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -83,7 +83,7 @@ ;; check. However, we need default value of SEW for vsetvl instruction since there ;; is no field for ratio in the vsetvl instruction encoding. (define_attr "sew" "" - (cond [(eq_attr "mode" "RVVMF64BI,RVVMF32BI,RVVMF16BI,RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ + (cond [(eq_attr "mode" "RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ RVVM8QI,RVVM4QI,RVVM2QI,RVVM1QI,RVVMF2QI,RVVMF4QI,RVVMF8QI,\ RVVM1x8QI,RVVMF2x8QI,RVVMF4x8QI,RVVMF8x8QI,\ RVVM1x7QI,RVVMF2x7QI,RVVMF4x7QI,RVVMF8x7QI,\ @@ -95,6 +95,18 @@ V1QI,V2QI,V4QI,V8QI,V16QI,V32QI,V64QI,V128QI,V256QI,V512QI,V1024QI,V2048QI,V4096QI,\ V1BI,V2BI,V4BI,V8BI,V16BI,V32BI,V64BI,V128BI,V256BI,V512BI,V1024BI,V2048BI,V4096BI") (const_int 8) + (eq_attr "mode" "RVVMF16BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 16) + (const_int 8)) + (eq_attr "mode" "RVVMF32BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 32) + (const_int 8)) + (eq_attr "mode" "RVVMF64BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 64) + (const_int 8)) (eq_attr "mode" "RVVM8HI,RVVM4HI,RVVM2HI,RVVM1HI,RVVMF2HI,RVVMF4HI,\ RVVM1x8HI,RVVMF2x8HI,RVVMF4x8HI,\ RVVM1x7HI,RVVMF2x7HI,RVVMF4x7HI,\ @@ -155,9 +167,9 @@ (eq_attr "mode" "RVVM4QI,RVVMF2BI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2QI,RVVMF4BI") (symbol_ref "riscv_vector::LMUL_2") (eq_attr "mode" "RVVM1QI,RVVMF8BI") (symbol_ref "riscv_vector::LMUL_1") - (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "riscv_vector::LMUL_F2") - (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "riscv_vector::LMUL_F4") - (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "riscv_vector::LMUL_F8") + (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F2") + (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F4") + (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F8") (eq_attr "mode" "RVVM8HI") (symbol_ref "riscv_vector::LMUL_8") (eq_attr "mode" "RVVM4HI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2HI") (symbol_ref "riscv_vector::LMUL_2") @@ -428,6 +440,10 @@ vislide1up,vislide1down,vfslide1up,vfslide1down,\ vgather,vcompress,vlsegdux,vlsegdox,vssegtux,vssegtox") (const_int INVALID_ATTRIBUTE) + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE) (eq_attr "mode" "RVVM8QI,RVVM1BI") (const_int 1) (eq_attr "mode" "RVVM4QI,RVVMF2BI") (const_int 2) (eq_attr "mode" "RVVM2QI,RVVMF4BI") (const_int 4) @@ -888,6 +904,8 @@ (symbol_ref "riscv_vector::FRM_DYN")] (symbol_ref "riscv_vector::FRM_NONE"))) +(include "thead-vector.md") + ;; ----------------------------------------------------------------- ;; ---- Miscellaneous Operations ;; ----------------------------------------------------------------- @@ -1097,7 +1115,7 @@ (define_insn "*mov<mode>_whole" [(set (match_operand:V_WHOLE 0 "reg_or_mem_operand" "=vr, m,vr") (match_operand:V_WHOLE 1 "reg_or_mem_operand" " m,vr,vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "@ vl%m1re<sew>.v\t%0,%1 vs%m1r.v\t%1,%0 @@ -1125,7 +1143,7 @@ (define_insn "*mov<mode>" [(set (match_operand:VB 0 "register_operand" "=vr") (match_operand:VB 1 "register_operand" " vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "vmv1r.v\t%0,%1" [(set_attr "type" "vmov") (set_attr "mode" "<MODE>")]) @@ -3680,7 +3698,7 @@ (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3701,7 +3719,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3722,7 +3740,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c index 2e0e12aa045..2eef9e1e1a8 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c @@ -1,4 +1,4 @@ -/* { dg-do compile } */ +/* { dg-do compile { target { ! riscv_xtheadvector } } } */ /* { dg-skip-if "test rvv intrinsic" { *-*-* } { "*" } { "-march=rv*v*" } } */ void foo0 () {__rvv_bool64_t t;} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c index 3d81b179235..ef329e30785 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c @@ -1,4 +1,4 @@ /* { dg-do compile } */ /* { dg-options "-O3 -march=rv32gc -mabi=ilp32d" } */ -#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' extension enabled} } */ +#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' or 'XTHEADVECTOR' extension enabled} } */ diff --git a/gcc/testsuite/lib/target-supports.exp b/gcc/testsuite/lib/target-supports.exp index 7f13ff0ca56..70df6b1401c 100644 --- a/gcc/testsuite/lib/target-supports.exp +++ b/gcc/testsuite/lib/target-supports.exp @@ -1952,6 +1952,18 @@ proc check_effective_target_riscv_zbb { } { }] } +# Return 1 if the target arch supports the XTheadVector extension, 0 otherwise. +# Cache the result. + +proc check_effective_target_riscv_xtheadvector { } { + return [check_no_compiler_messages riscv_ext_xtheadvector assembly { + #ifndef __riscv_xtheadvector + #error "Not __riscv_xtheadvector" + #endif + }] +} + + # Return 1 if we can execute code when using dg-add-options riscv_v proc check_effective_target_riscv_v_ok { } { -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re：[PATCH v4] RISC-V: Handle differences between XTheadvector and Vector 2024-01-02 2:00 ` juzhe.zhong 2024-01-02 3:03 ` Re：[PATCH " joshua @ 2024-01-02 9:48 ` joshua 2024-01-02 9:52 ` juzhe.zhong 1 sibling, 1 reply; 133+ messages in thread From: joshua @ 2024-01-02 9:48 UTC (permalink / raw) To: juzhe.zhong, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (mode, dest, src, + RVV_VLMAX, GEN_INT(VLMAX))); + return true; + } Move it outside legitimize_move It should be it: if (TARGET_THEADVECTOR) emit_th_move... DONE; else if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; If we move emit_insn (gen_pred_th_whole_mov...) outside legitimize, we need to modify the patterns which will call legitimize_move (). I have tried and that will result in 5 extra changes in vector.md. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月2日(星期二) 10:00 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v4] RISC-V: Handle differences between XTheadvector and Vector + if (TARGET_XTHEADVECTOR) + return false; Move it to : if (TARGET_VECTOR && stringop_strategy & STRATEGY_VECTOR) { bool ok = riscv_vector::expand_block_move (dest, src, length); if (ok) return true; } (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (match_test "!TARGET_XTHEADVECTOR || rtx_equal_p (op, const0_rtx)") + (match_operand 0 "const_csr_operand")))) It's hard to trace. Change it into : (ior 1. TARGET_THEADVECTOR && rtx_equal_p (op, const0_rtx) 2. !TAGEET_THEADVECTOR && const_csr_operand) + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (mode, dest, src, + RVV_VLMAX, GEN_INT(VLMAX))); + return true; + } Move it outside legitimize_move It should be it: if (TARGET_THEADVECTOR) emit_th_move... DONE; else if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; vsetvli issues: I wonder whether we can use ASM_OUTPUT_OPCODE to recognize "ta,ma"/"ta,mu"/"tu,ma"/"tu,mu" and replace these 4 variants by "". So that we don't have tail policy and mask policy in vsetvli ASM string. Another alternative approach is we can change vsetlvi ASM rule: "vset%i1vli\t%0,%1,e%2,%m3,t%p4,m%p5" if (TARGET_THEADVECTOR) ... else else if (code == CONST_INT) { /* Tail && Mask policy. */ asm_fprintf (file, "%s", IS_AGNOSTIC (UINTVAL (op)) ? "a" : "u"); } in riscv.cc. The benefit is that we can avoid adding all th_vsetvl patterns and invasive code changs in VSETVL PASS. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2023-12-29 12:21 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v4] RISC-V: Handle differences between XTheadvector and Vector This patch is to handle the differences in instruction generation between Vector and XTheadVector. In this version, we only support partial xtheadvector instructions that leverage directly from current RVV1.0 with simple adding "th." prefix. For different name xtheadvector instructions but share same patterns as RVV1.0 instructions, we will use ASM targethook to rewrite the whole string of the instructions in the following patches. For some vector patterns that cannot be avoided, we use "!TARGET_XTHEADVECTOR" to disable them in vector.md in order not to generate instructions that xtheadvector does not support, like vmv1r and vsext.vf2. gcc/ChangeLog: * config.gcc: Add files for XTheadVector intrinsics. * config/riscv/autovec.md: Guard XTheadVector. * config/riscv/riscv-string.cc (expand_block_move): Guard XTheadVector. * config/riscv/riscv-v.cc (legitimize_move): New expansion. (get_prefer_tail_policy): Give specific value for tail. (get_prefer_mask_policy): Give specific value for mask. (vls_mode_valid_p): Avoid autovec. * config/riscv/riscv-vector-builtins-shapes.cc (check_type): (build_one): New function. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): (DEF_THEAD_RVV_FUNCTION): Add new marcos. (check_required_extensions): (handle_pragma_vector): * config/riscv/riscv-vector-builtins.h (RVV_REQUIRE_VECTOR): (RVV_REQUIRE_XTHEADVECTOR): Add RVV_REQUIRE_VECTOR and RVV_REQUIRE_XTHEADVECTOR. (struct function_group_info): * config/riscv/riscv-vector-switch.def (ENTRY): Disable fractional mode for the XTheadVector extension. (TUPLE_ENTRY): Likewise. * config/riscv/riscv-vsetvl.cc: Add functions for xtheadvector. * config/riscv/riscv.cc (riscv_v_ext_vls_mode_p): Guard XTheadVector. (riscv_v_adjust_bytesize): Likewise. (riscv_preferred_simd_mode): Likewsie. (riscv_autovectorize_vector_modes): Likewise. (riscv_vector_mode_supported_any_target_p): Likewise. (TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P): Likewise. * config/riscv/vector-iterators.md: Remove fractional LMUL. * config/riscv/vector.md: Include thead-vector.md. * config/riscv/riscv_th_vector.h: New file. * config/riscv/thead-vector.md: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/base/pragma-1.c: Add XTheadVector. * gcc.target/riscv/rvv/base/abi-1.c: Exclude XTheadVector. * lib/target-supports.exp: Add target for XTheadVector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config.gcc | 2 +- gcc/config/riscv/autovec.md | 2 +- gcc/config/riscv/predicates.md | 3 +- gcc/config/riscv/riscv-string.cc | 3 + gcc/config/riscv/riscv-v.cc | 13 +- .../riscv/riscv-vector-builtins-bases.cc | 3 + .../riscv/riscv-vector-builtins-shapes.cc | 23 +++ gcc/config/riscv/riscv-vector-switch.def | 150 +++++++------- gcc/config/riscv/riscv-vsetvl.cc | 10 + gcc/config/riscv/riscv.cc | 20 +- gcc/config/riscv/riscv_th_vector.h | 49 +++++ gcc/config/riscv/thead-vector.md | 142 +++++++++++++ gcc/config/riscv/vector-iterators.md | 186 +++++++++--------- gcc/config/riscv/vector.md | 36 +++- .../gcc.target/riscv/rvv/base/abi-1.c | 2 +- .../gcc.target/riscv/rvv/base/pragma-1.c | 2 +- gcc/testsuite/lib/target-supports.exp | 12 ++ 17 files changed, 471 insertions(+), 187 deletions(-) create mode 100644 gcc/config/riscv/riscv_th_vector.h create mode 100644 gcc/config/riscv/thead-vector.md diff --git a/gcc/config.gcc b/gcc/config.gcc index f0676c830e8..1445d98c147 100644 --- a/gcc/config.gcc +++ b/gcc/config.gcc @@ -549,7 +549,7 @@ riscv*) extra_objs="${extra_objs} riscv-vector-builtins.o riscv-vector-builtins-shapes.o riscv-vector-builtins-bases.o" extra_objs="${extra_objs} thead.o riscv-target-attr.o" d_target_objs="riscv-d.o" - extra_headers="riscv_vector.h" + extra_headers="riscv_vector.h riscv_th_vector.h" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.cc" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.h" ;; diff --git a/gcc/config/riscv/autovec.md b/gcc/config/riscv/autovec.md index 8b8a92f10a1..1fac56c7095 100644 --- a/gcc/config/riscv/autovec.md +++ b/gcc/config/riscv/autovec.md @@ -2579,7 +2579,7 @@ [(match_operand 0 "register_operand") (match_operand 1 "memory_operand") (match_operand:ANYI 2 "const_int_operand")] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" { riscv_vector::expand_rawmemchr(<MODE>mode, operands[0], operands[1], operands[2]); diff --git a/gcc/config/riscv/predicates.md b/gcc/config/riscv/predicates.md index b87a6900841..c5ab86d5db9 100644 --- a/gcc/config/riscv/predicates.md +++ b/gcc/config/riscv/predicates.md @@ -428,7 +428,8 @@ ;; Predicates for the V extension. (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (match_test "!TARGET_XTHEADVECTOR || rtx_equal_p (op, const0_rtx)") + (match_operand 0 "const_csr_operand")))) (define_special_predicate "autovec_length_operand" (ior (match_operand 0 "pmode_register_operand") diff --git a/gcc/config/riscv/riscv-string.cc b/gcc/config/riscv/riscv-string.cc index 11c1f74d0b3..ec8f3486fd8 100644 --- a/gcc/config/riscv/riscv-string.cc +++ b/gcc/config/riscv/riscv-string.cc @@ -808,6 +808,9 @@ expand_block_move (rtx dst_in, rtx src_in, rtx length_in) bnez a2, loop # Any more? ret # Return */ + if (TARGET_XTHEADVECTOR) + return false; + gcc_assert (TARGET_VECTOR); HOST_WIDE_INT potential_ew diff --git a/gcc/config/riscv/riscv-v.cc b/gcc/config/riscv/riscv-v.cc index 038ab084a37..5e9e45aecd2 100644 --- a/gcc/config/riscv/riscv-v.cc +++ b/gcc/config/riscv/riscv-v.cc @@ -1523,6 +1523,13 @@ legitimize_move (rtx dest, rtx *srcp) return true; } + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (mode, dest, src, + RVV_VLMAX, GEN_INT(VLMAX))); + return true; + } + if (riscv_v_ext_vls_mode_p (mode)) { if (GET_MODE_NUNITS (mode).to_constant () <= 31) @@ -1772,7 +1779,7 @@ get_prefer_tail_policy () compiler pick up either agnostic or undisturbed. Maybe we will have a compile option like -mprefer=agnostic to set this value???. */ - return TAIL_ANY; + return TARGET_XTHEADVECTOR ? TAIL_AGNOSTIC : TAIL_ANY; } /* Get prefer mask policy. */ @@ -1783,7 +1790,7 @@ get_prefer_mask_policy () compiler pick up either agnostic or undisturbed. Maybe we will have a compile option like -mprefer=agnostic to set this value???. */ - return MASK_ANY; + return TARGET_XTHEADVECTOR ? MASK_UNDISTURBED : MASK_ANY; } /* Get avl_type rtx. */ @@ -4383,7 +4390,7 @@ cmp_lmul_gt_one (machine_mode mode) bool vls_mode_valid_p (machine_mode vls_mode) { - if (!TARGET_VECTOR) + if (!TARGET_VECTOR || TARGET_XTHEADVECTOR) return false; if (riscv_autovec_preference == RVV_SCALABLE) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index c51affde353..2918c07ebf3 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -133,6 +133,9 @@ public: = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + if (TARGET_XTHEADVECTOR) + return e.generate_insn (code_for_th_vsetvl_no_side_effects (Pmode)); + /* TAIL_ANY. */ e.add_input_operand (Pmode, gen_int_mode (get_prefer_tail_policy (), Pmode)); diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 4a754e0228f..6b49404a1fa 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -33,6 +33,25 @@ namespace riscv_vector { +/* Check whether the RETURN_TYPE and ARGUMENT_TYPES are + valid for the function. */ + +static bool +check_type (tree return_type, vec<tree> &argument_types) +{ + tree arg; + unsigned i; + + if (!return_type) + return false; + + FOR_EACH_VEC_ELT (argument_types, i, arg) + if (!arg) + return false; + + return true; +} + /* Add one function instance for GROUP, using operand suffix at index OI, mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ static void @@ -49,6 +68,10 @@ build_one (function_builder &b, const function_group_info &group, group.ops_infos.types[vec_type_idx].index); b.allocate_argument_types (function_instance, argument_types); b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + b.add_overloaded_function (function_instance, *group.shape); b.add_unique_function (function_instance, (*group.shape), return_type, argument_types); diff --git a/gcc/config/riscv/riscv-vector-switch.def b/gcc/config/riscv/riscv-vector-switch.def index 5c9f9bcbc3e..f7a66b34bae 100644 --- a/gcc/config/riscv/riscv-vector-switch.def +++ b/gcc/config/riscv/riscv-vector-switch.def @@ -68,9 +68,9 @@ Encode the ratio of SEW/LMUL into the mask types. #endif /* Disable modes if TARGET_MIN_VLEN == 32. */ -ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) -ENTRY (RVVMF32BI, true, LMUL_F4, 32) -ENTRY (RVVMF16BI, true, LMUL_F2, 16) +ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F8, 64) +ENTRY (RVVMF32BI, true, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F4, 32) +ENTRY (RVVMF16BI, true, TARGET_XTHEADVECTOR ? LMUL_1 : LMUL_F2 , 16) ENTRY (RVVMF8BI, true, LMUL_1, 8) ENTRY (RVVMF4BI, true, LMUL_2, 4) ENTRY (RVVMF2BI, true, LMUL_4, 2) @@ -81,39 +81,39 @@ ENTRY (RVVM8QI, true, LMUL_8, 1) ENTRY (RVVM4QI, true, LMUL_4, 2) ENTRY (RVVM2QI, true, LMUL_2, 4) ENTRY (RVVM1QI, true, LMUL_1, 8) -ENTRY (RVVMF2QI, true, LMUL_F2, 16) -ENTRY (RVVMF4QI, true, LMUL_F4, 32) -ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) +ENTRY (RVVMF2QI, !TARGET_XTHEADVECTOR, LMUL_F2, 16) +ENTRY (RVVMF4QI, !TARGET_XTHEADVECTOR, LMUL_F4, 32) +ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F8, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8HI, true, LMUL_8, 2) ENTRY (RVVM4HI, true, LMUL_4, 4) ENTRY (RVVM2HI, true, LMUL_2, 8) ENTRY (RVVM1HI, true, LMUL_1, 16) -ENTRY (RVVMF2HI, true, LMUL_F2, 32) -ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HI, !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_16. */ ENTRY (RVVM8HF, TARGET_VECTOR_ELEN_FP_16, LMUL_8, 2) ENTRY (RVVM4HF, TARGET_VECTOR_ELEN_FP_16, LMUL_4, 4) ENTRY (RVVM2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_2, 8) ENTRY (RVVM1HF, TARGET_VECTOR_ELEN_FP_16, LMUL_1, 16) -ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_F2, 32) -ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8SI, true, LMUL_8, 4) ENTRY (RVVM4SI, true, LMUL_4, 8) ENTRY (RVVM2SI, true, LMUL_2, 16) ENTRY (RVVM1SI, true, LMUL_1, 32) -ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_32. */ ENTRY (RVVM8SF, TARGET_VECTOR_ELEN_FP_32, LMUL_8, 4) ENTRY (RVVM4SF, TARGET_VECTOR_ELEN_FP_32, LMUL_4, 8) ENTRY (RVVM2SF, TARGET_VECTOR_ELEN_FP_32, LMUL_2, 16) ENTRY (RVVM1SF, TARGET_VECTOR_ELEN_FP_32, LMUL_1, 32) -ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if !TARGET_VECTOR_ELEN_64. */ ENTRY (RVVM8DI, TARGET_VECTOR_ELEN_64, LMUL_8, 8) @@ -140,127 +140,127 @@ ENTRY (RVVM1DF, TARGET_VECTOR_ELEN_FP_64, LMUL_1, 64) #endif TUPLE_ENTRY (RVVM1x8QI, true, RVVM1QI, 8, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x8QI, true, RVVMF2QI, 8, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x8QI, true, RVVMF4QI, 8, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 8, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x8QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 8, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x8QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 8, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 8, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x7QI, true, RVVM1QI, 7, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x7QI, true, RVVMF2QI, 7, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x7QI, true, RVVMF4QI, 7, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 7, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x7QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 7, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x7QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 7, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 7, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x6QI, true, RVVM1QI, 6, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x6QI, true, RVVMF2QI, 6, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x6QI, true, RVVMF4QI, 6, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 6, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x6QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 6, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x6QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 6, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 6, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x5QI, true, RVVM1QI, 5, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x5QI, true, RVVMF2QI, 5, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x5QI, true, RVVMF4QI, 5, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 5, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x5QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 5, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x5QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 5, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 5, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x4QI, true, RVVM2QI, 4, LMUL_2, 4) TUPLE_ENTRY (RVVM1x4QI, true, RVVM1QI, 4, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x4QI, true, RVVMF2QI, 4, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x4QI, true, RVVMF4QI, 4, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 4, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x4QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 4, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x4QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 4, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 4, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x3QI, true, RVVM2QI, 3, LMUL_2, 4) TUPLE_ENTRY (RVVM1x3QI, true, RVVM1QI, 3, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x3QI, true, RVVMF2QI, 3, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x3QI, true, RVVMF4QI, 3, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 3, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x3QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 3, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x3QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 3, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 3, LMUL_F8, 64) TUPLE_ENTRY (RVVM4x2QI, true, RVVM4QI, 2, LMUL_4, 2) TUPLE_ENTRY (RVVM2x2QI, true, RVVM2QI, 2, LMUL_2, 4) TUPLE_ENTRY (RVVM1x2QI, true, RVVM1QI, 2, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x2QI, true, RVVMF2QI, 2, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x2QI, true, RVVMF4QI, 2, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 2, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x2QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 2, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x2QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 2, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 2, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x8HI, true, RVVM1HI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HI, true, RVVMF2HI, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HI, true, RVVM1HI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HI, true, RVVMF2HI, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HI, true, RVVM1HI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HI, true, RVVMF2HI, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HI, true, RVVM1HI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HI, true, RVVMF2HI, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HI, true, RVVM2HI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HI, true, RVVM1HI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HI, true, RVVMF2HI, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HI, true, RVVM2HI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HI, true, RVVM1HI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HI, true, RVVMF2HI, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HI, true, RVVM4HI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HI, true, RVVM2HI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HI, true, RVVM1HI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HI, true, RVVMF2HI, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM4HF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8SI, true, RVVM1SI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SI, true, RVVM1SI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SI, true, RVVM1SI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SI, true, RVVM1SI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SI, true, RVVM2SI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SI, true, RVVM1SI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SI, true, RVVM2SI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SI, true, RVVM1SI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SI, true, RVVM4SI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SI, true, RVVM2SI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SI, true, RVVM1SI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM4SF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 8, LMUL_1, 16) TUPLE_ENTRY (RVVM1x7DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 7, LMUL_1, 16) diff --git a/gcc/config/riscv/riscv-vsetvl.cc b/gcc/config/riscv/riscv-vsetvl.cc index eabaef80f89..c726253c107 100644 --- a/gcc/config/riscv/riscv-vsetvl.cc +++ b/gcc/config/riscv/riscv-vsetvl.cc @@ -1117,6 +1117,16 @@ public: avl = GEN_INT (0); rtx sew = gen_int_mode (get_sew (), Pmode); rtx vlmul = gen_int_mode (get_vlmul (), Pmode); + + if (TARGET_XTHEADVECTOR) { + if (change_vtype_only_p ()) + return gen_th_vsetvl_vtype_change_only (sew, vlmul); + else if (has_vl () && !ignore_vl) + return gen_th_vsetvl (Pmode, get_vl (), avl, sew, vlmul); + else + return gen_th_vsetvl_discard_result (Pmode, avl, sew, vlmul); + } + rtx ta = gen_int_mode (get_ta (), Pmode); rtx ma = gen_int_mode (get_ma (), Pmode); diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index ea1d59d9cf2..a80bf8d1a74 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -1406,6 +1406,9 @@ riscv_v_adjust_bytesize (machine_mode mode, int scale) { if (riscv_v_ext_vector_mode_p (mode)) { + if (TARGET_XTHEADVECTOR) + return BYTES_PER_RISCV_VECTOR; + poly_int64 nunits = GET_MODE_NUNITS (mode); poly_int64 mode_size = GET_MODE_SIZE (mode); @@ -9965,7 +9968,7 @@ riscv_use_divmod_expander (void) static machine_mode riscv_preferred_simd_mode (scalar_mode mode) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::preferred_simd_mode (mode); return word_mode; @@ -10316,7 +10319,7 @@ riscv_mode_priority (int, int n) unsigned int riscv_autovectorize_vector_modes (vector_modes *modes, bool all) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::autovectorize_vector_modes (modes, all); return default_autovectorize_vector_modes (modes, all); @@ -10499,6 +10502,16 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) return false; } +/* Implements target hook vector_mode_supported_any_target_p. */ + +static bool +riscv_vector_mode_supported_any_target_p (machine_mode mode) +{ + if (TARGET_XTHEADVECTOR) + return false; + return true; +} + /* Initialize the GCC target structure. */ #undef TARGET_ASM_ALIGNED_HI_OP #define TARGET_ASM_ALIGNED_HI_OP "\t.half\t" @@ -10842,6 +10855,9 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) #undef TARGET_PREFERRED_ELSE_VALUE #define TARGET_PREFERRED_ELSE_VALUE riscv_preferred_else_value +#undef TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P +#define TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P riscv_vector_mode_supported_any_target_p + struct gcc_target targetm = TARGET_INITIALIZER; #include "gt-riscv.h" diff --git a/gcc/config/riscv/riscv_th_vector.h b/gcc/config/riscv/riscv_th_vector.h new file mode 100644 index 00000000000..6f47e0c90a4 --- /dev/null +++ b/gcc/config/riscv/riscv_th_vector.h @@ -0,0 +1,49 @@ +/* RISC-V 'XTheadVector' Extension intrinsics include file. + Copyright (C) 2022-2023 Free Software Foundation, Inc. + + This file is part of GCC. + + GCC is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published + by the Free Software Foundation; either version 3, or (at your + option) any later version. + + GCC is distributed in the hope that it will be useful, but WITHOUT + ANY WARRANTY; without even the implied warranty of MERCHANTABILITY + or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public + License for more details. + + Under Section 7 of GPL version 3, you are granted additional + permissions described in the GCC Runtime Library Exception, version + 3.1, as published by the Free Software Foundation. + + You should have received a copy of the GNU General Public License and + a copy of the GCC Runtime Library Exception along with this program; + see the files COPYING3 and COPYING.RUNTIME respectively. If not, see + <" target="_blank">http://www.gnu.org/licenses/>. */ + +#ifndef __RISCV_TH_VECTOR_H +#define __RISCV_TH_VECTOR_H + +#include <stdint.h> +#include <stddef.h> + +#ifndef __riscv_xtheadvector +#error "XTheadVector intrinsics require the xtheadvector extension." +#else +#ifdef __cplusplus +extern "C" { +#endif + +/* NOTE: This implementation of riscv_th_vector.h is intentionally short. It does + not define the RVV types and intrinsic functions directly in C and C++ + code, but instead uses the following pragma to tell GCC to insert the + necessary type and function definitions itself. The net effect is the + same, and the file is a complete implementation of riscv_th_vector.h. */ +#pragma riscv intrinsic "vector" + +#ifdef __cplusplus +} +#endif // __cplusplus +#endif // __riscv_xtheadvector +#endif // __RISCV_TH_ECTOR_H diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md new file mode 100644 index 00000000000..f0aa196d93b --- /dev/null +++ b/gcc/config/riscv/thead-vector.md @@ -0,0 +1,142 @@ +(define_c_enum "unspec" [ + UNSPEC_TH_VWLDST +]) + +(define_mode_iterator V_VLS_VT [V VLS VT]) +(define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) + +(define_split + [(set (match_operand:V_VB_VLS_VT 0 "reg_or_mem_operand") + (match_operand:V_VB_VLS_VT 1 "reg_or_mem_operand"))] + "TARGET_XTHEADVECTOR" + [(const_int 0)] + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + }) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:V_VLS_VT 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:V_VLS_VT + [(match_operand:V_VLS_VT 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2")]) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:VB 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:VB + [(match_operand:VB 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2") + (set (attr "sew") (const_int 8)) + (set (attr "vlmul") (symbol_ref "riscv_vector::LMUL_1"))]) + +(define_insn "@th_vsetvl<mode>" + [(set (match_operand:P 0 "register_operand" "=r") + (unspec:P [(match_operand:P 1 "vector_length_operand" "rK") + (match_operand 2 "const_int_operand" "i") + (match_operand 3 "const_int_operand" "i")] UNSPEC_VSETVL)) + (set (reg:SI VL_REGNUM) + (unspec:SI [(match_dup 1) + (match_dup 2) + (match_dup 3)] UNSPEC_VSETVL)) + (set (reg:SI VTYPE_REGNUM) + (unspec:SI [(match_dup 2) + (match_dup 3)] UNSPEC_VSETVL))] + "TARGET_XTHEADVECTOR" + "vsetvli\t%0,%1,e%2,%m3" + [(set_attr "type" "vsetvl") + (set_attr "mode" "<MODE>") + (set (attr "sew") (symbol_ref "INTVAL (operands[2])")) + (set (attr "vlmul") (symbol_ref "INTVAL (operands[3])"))]) + +;; vsetvl zero,zero,vtype instruction. +;; This pattern has no side effects and does not set X0 register. +(define_insn "th_vsetvl_vtype_change_only" + [(set (reg:SI VTYPE_REGNUM) + (unspec:SI + [(match_operand 0 "const_int_operand" "i") + (match_operand 1 "const_int_operand" "i")] UNSPEC_VSETVL))] + "TARGET_XTHEADVECTOR" + "vsetvli\tzero,zero,e%0,%m1" + [(set_attr "type" "vsetvl") + (set_attr "mode" "SI") + (set (attr "sew") (symbol_ref "INTVAL (operands[0])")) + (set (attr "vlmul") (symbol_ref "INTVAL (operands[1])"))]) + +;; vsetvl zero,rs1,vtype instruction. +;; The reason we need this pattern since we should avoid setting X0 register +;; in vsetvl instruction pattern. +(define_insn "@th_vsetvl_discard_result<mode>" + [(set (reg:SI VL_REGNUM) + (unspec:SI [(match_operand:P 0 "vector_length_operand" "rK") + (match_operand 1 "const_int_operand" "i") + (match_operand 2 "const_int_operand" "i")] UNSPEC_VSETVL)) + (set (reg:SI VTYPE_REGNUM) + (unspec:SI [(match_dup 1) + (match_dup 2)] UNSPEC_VSETVL))] + "TARGET_XTHEADVECTOR" + "vsetvli\tzero,%0,e%1,%m2" + [(set_attr "type" "vsetvl") + (set_attr "mode" "<MODE>") + (set (attr "sew") (symbol_ref "INTVAL (operands[1])")) + (set (attr "vlmul") (symbol_ref "INTVAL (operands[2])"))]) + +;; It's emit by vsetvl/vsetvlmax intrinsics with no side effects. +;; Since we have many optmization passes from "expand" to "reload_completed", +;; such pattern can allow us gain benefits of these optimizations. +(define_insn_and_split "@th_vsetvl<mode>_no_side_effects" + [(set (match_operand:P 0 "register_operand" "=r") + (unspec:P [(match_operand:P 1 "vector_length_operand" "rK") + (match_operand 2 "const_int_operand" "i") + (match_operand 3 "const_int_operand" "i")] UNSPEC_VSETVL))] + "TARGET_XTHEADVECTOR" + "#" + "&& epilogue_completed" + [(parallel + [(set (match_dup 0) + (unspec:P [(match_dup 1) (match_dup 2) (match_dup 3)] UNSPEC_VSETVL)) + (set (reg:SI VL_REGNUM) + (unspec:SI [(match_dup 1) (match_dup 2) (match_dup 3)] UNSPEC_VSETVL)) + (set (reg:SI VTYPE_REGNUM) + (unspec:SI [(match_dup 2) (match_dup 3)] UNSPEC_VSETVL))])] + "" + [(set_attr "type" "vsetvl") + (set_attr "mode" "SI")]) diff --git a/gcc/config/riscv/vector-iterators.md b/gcc/config/riscv/vector-iterators.md index 5f5f7b5b986..c0fc7a2441d 100644 --- a/gcc/config/riscv/vector-iterators.md +++ b/gcc/config/riscv/vector-iterators.md @@ -109,11 +109,11 @@ ]) (define_mode_iterator VI [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -128,11 +128,11 @@ ;; allow the instruction and mode to be matched during combine et al. (define_mode_iterator VF [ (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -140,11 +140,11 @@ (define_mode_iterator VF_ZVFHMIN [ (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -271,16 +271,16 @@ ]) (define_mode_iterator VEEWEXT2 [ - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -290,10 +290,10 @@ ]) (define_mode_iterator VEEWEXT4 [ - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -311,59 +311,59 @@ ]) (define_mode_iterator VEEWTRUNC2 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") (RVVM4SI "TARGET_64BIT") (RVVM2SI "TARGET_64BIT") (RVVM1SI "TARGET_64BIT") - (RVVMF2SI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEEWTRUNC4 [ - RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM2HI "TARGET_64BIT") (RVVM1HI "TARGET_64BIT") - (RVVMF2HI "TARGET_64BIT") - (RVVMF4HI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2HI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") (RVVM1HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEEWTRUNC8 [ (RVVM1QI "TARGET_64BIT") - (RVVMF2QI "TARGET_64BIT") - (RVVMF4QI "TARGET_64BIT") - (RVVMF8QI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2QI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF4QI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEI16 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -452,11 +452,11 @@ ]) (define_mode_iterator VFULLI [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_FULL_V") (RVVM4DI "TARGET_FULL_V") (RVVM2DI "TARGET_FULL_V") (RVVM1DI "TARGET_FULL_V") @@ -509,17 +509,17 @@ ]) (define_mode_iterator VI_QH [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") ]) (define_mode_iterator VI_QHS [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -560,11 +560,11 @@ ]) (define_mode_iterator VI_QHS_NO_M8 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -603,11 +603,11 @@ (define_mode_iterator VF_HS [ (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (V1HF "riscv_vector::vls_mode_valid_p (V1HFmode) && TARGET_ZVFH") (V2HF "riscv_vector::vls_mode_valid_p (V2HFmode) && TARGET_ZVFH") @@ -638,12 +638,12 @@ (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") - (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (V1HF "riscv_vector::vls_mode_valid_p (V1HFmode) && TARGET_ZVFH") (V2HF "riscv_vector::vls_mode_valid_p (V2HFmode) && TARGET_ZVFH") @@ -674,11 +674,11 @@ ]) (define_mode_iterator V_VLSI_QHS [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -756,27 +756,27 @@ ;; E.g. when index mode = RVVM8QImde and Pmode = SImode, if it is not zero_extend or ;; scalar != 1, such gather/scatter is not allowed since we don't have RVVM32SImode. (define_mode_iterator RATIO64 [ - (RVVMF8QI "TARGET_MIN_VLEN > 32") - (RVVMF4HI "TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM1DI "TARGET_VECTOR_ELEN_64") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") ]) (define_mode_iterator RATIO32 [ - RVVMF4QI - RVVMF2HI + (RVVMF4QI "!TARGET_XTHEADVECTOR") + (RVVMF2HI "!TARGET_XTHEADVECTOR") RVVM1SI (RVVM2DI "TARGET_VECTOR_ELEN_64") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") ]) (define_mode_iterator RATIO16 [ - RVVMF2QI + (RVVMF2QI "!TARGET_XTHEADVECTOR") RVVM1HI RVVM2SI (RVVM4DI "TARGET_VECTOR_ELEN_64") @@ -814,21 +814,21 @@ ]) (define_mode_iterator RATIO64I [ - (RVVMF8QI "TARGET_MIN_VLEN > 32") - (RVVMF4HI "TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM1DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") ]) (define_mode_iterator RATIO32I [ - RVVMF4QI - RVVMF2HI + (RVVMF4QI "!TARGET_XTHEADVECTOR") + (RVVMF2HI "!TARGET_XTHEADVECTOR") RVVM1SI (RVVM2DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") ]) (define_mode_iterator RATIO16I [ - RVVMF2QI + (RVVMF2QI "!TARGET_XTHEADVECTOR") RVVM1HI RVVM2SI (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") @@ -873,21 +873,21 @@ ]) (define_mode_iterator V_FRACT [ - RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") ]) (define_mode_iterator VWEXTI [ - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -933,7 +933,7 @@ (RVVM4SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -966,7 +966,7 @@ (RVVM4SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -996,7 +996,7 @@ (define_mode_iterator VWCONVERTI [ (RVVM8SI "TARGET_ZVFH") (RVVM4SI "TARGET_ZVFH") (RVVM2SI "TARGET_ZVFH") (RVVM1SI "TARGET_ZVFH") - (RVVMF2SI "TARGET_ZVFH") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_ZVFH") (RVVM8DI "TARGET_VECTOR_ELEN_64 && TARGET_VECTOR_ELEN_FP_32") (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_VECTOR_ELEN_FP_32") @@ -1045,7 +1045,7 @@ ]) (define_mode_iterator VQEXTI [ - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -1456,11 +1456,11 @@ ;; VINDEXED [VI8 VI16 VI32 (VI64 "TARGET_64BIT")]. (define_mode_iterator VINDEXED [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") @@ -1468,12 +1468,12 @@ (RVVM1DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64 && TARGET_64BIT") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64 && TARGET_64BIT") @@ -3173,11 +3173,11 @@ (define_mode_iterator V_VLS_F_CONVERT_SI [ (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") - (RVVMF2HF "TARGET_ZVFH") (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") @@ -3290,12 +3290,12 @@ ]) (define_mode_iterator V_VLS_F_CONVERT_DI [ - (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index b5a9055cdc4..9e7a487759a 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -83,7 +83,7 @@ ;; check. However, we need default value of SEW for vsetvl instruction since there ;; is no field for ratio in the vsetvl instruction encoding. (define_attr "sew" "" - (cond [(eq_attr "mode" "RVVMF64BI,RVVMF32BI,RVVMF16BI,RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ + (cond [(eq_attr "mode" "RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ RVVM8QI,RVVM4QI,RVVM2QI,RVVM1QI,RVVMF2QI,RVVMF4QI,RVVMF8QI,\ RVVM1x8QI,RVVMF2x8QI,RVVMF4x8QI,RVVMF8x8QI,\ RVVM1x7QI,RVVMF2x7QI,RVVMF4x7QI,RVVMF8x7QI,\ @@ -95,6 +95,18 @@ V1QI,V2QI,V4QI,V8QI,V16QI,V32QI,V64QI,V128QI,V256QI,V512QI,V1024QI,V2048QI,V4096QI,\ V1BI,V2BI,V4BI,V8BI,V16BI,V32BI,V64BI,V128BI,V256BI,V512BI,V1024BI,V2048BI,V4096BI") (const_int 8) + (eq_attr "mode" "RVVMF16BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 16) + (const_int 8)) + (eq_attr "mode" "RVVMF32BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 32) + (const_int 8)) + (eq_attr "mode" "RVVMF64BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 64) + (const_int 8)) (eq_attr "mode" "RVVM8HI,RVVM4HI,RVVM2HI,RVVM1HI,RVVMF2HI,RVVMF4HI,\ RVVM1x8HI,RVVMF2x8HI,RVVMF4x8HI,\ RVVM1x7HI,RVVMF2x7HI,RVVMF4x7HI,\ @@ -155,9 +167,9 @@ (eq_attr "mode" "RVVM4QI,RVVMF2BI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2QI,RVVMF4BI") (symbol_ref "riscv_vector::LMUL_2") (eq_attr "mode" "RVVM1QI,RVVMF8BI") (symbol_ref "riscv_vector::LMUL_1") - (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "riscv_vector::LMUL_F2") - (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "riscv_vector::LMUL_F4") - (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "riscv_vector::LMUL_F8") + (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F2") + (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F4") + (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F8") (eq_attr "mode" "RVVM8HI") (symbol_ref "riscv_vector::LMUL_8") (eq_attr "mode" "RVVM4HI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2HI") (symbol_ref "riscv_vector::LMUL_2") @@ -428,6 +440,10 @@ vislide1up,vislide1down,vfslide1up,vfslide1down,\ vgather,vcompress,vlsegdux,vlsegdox,vssegtux,vssegtox") (const_int INVALID_ATTRIBUTE) + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE) (eq_attr "mode" "RVVM8QI,RVVM1BI") (const_int 1) (eq_attr "mode" "RVVM4QI,RVVMF2BI") (const_int 2) (eq_attr "mode" "RVVM2QI,RVVMF4BI") (const_int 4) @@ -888,6 +904,8 @@ (symbol_ref "riscv_vector::FRM_DYN")] (symbol_ref "riscv_vector::FRM_NONE"))) +(include "thead-vector.md") + ;; ----------------------------------------------------------------- ;; ---- Miscellaneous Operations ;; ----------------------------------------------------------------- @@ -1097,7 +1115,7 @@ (define_insn "*mov<mode>_whole" [(set (match_operand:V_WHOLE 0 "reg_or_mem_operand" "=vr, m,vr") (match_operand:V_WHOLE 1 "reg_or_mem_operand" " m,vr,vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "@ vl%m1re<sew>.v\t%0,%1 vs%m1r.v\t%1,%0 @@ -1125,7 +1143,7 @@ (define_insn "*mov<mode>" [(set (match_operand:VB 0 "register_operand" "=vr") (match_operand:VB 1 "register_operand" " vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "vmv1r.v\t%0,%1" [(set_attr "type" "vmov") (set_attr "mode" "<MODE>")]) @@ -3680,7 +3698,7 @@ (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3701,7 +3719,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3722,7 +3740,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c index 2e0e12aa045..2eef9e1e1a8 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c @@ -1,4 +1,4 @@ -/* { dg-do compile } */ +/* { dg-do compile { target { ! riscv_xtheadvector } } } */ /* { dg-skip-if "test rvv intrinsic" { *-*-* } { "*" } { "-march=rv*v*" } } */ void foo0 () {__rvv_bool64_t t;} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c index 3d81b179235..ef329e30785 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c @@ -1,4 +1,4 @@ /* { dg-do compile } */ /* { dg-options "-O3 -march=rv32gc -mabi=ilp32d" } */ -#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' extension enabled} } */ +#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' or 'XTHEADVECTOR' extension enabled} } */ diff --git a/gcc/testsuite/lib/target-supports.exp b/gcc/testsuite/lib/target-supports.exp index 7f13ff0ca56..70df6b1401c 100644 --- a/gcc/testsuite/lib/target-supports.exp +++ b/gcc/testsuite/lib/target-supports.exp @@ -1952,6 +1952,18 @@ proc check_effective_target_riscv_zbb { } { }] } +# Return 1 if the target arch supports the XTheadVector extension, 0 otherwise. +# Cache the result. + +proc check_effective_target_riscv_xtheadvector { } { + return [check_no_compiler_messages riscv_ext_xtheadvector assembly { + #ifndef __riscv_xtheadvector + #error "Not __riscv_xtheadvector" + #endif + }] +} + + # Return 1 if we can execute code when using dg-add-options riscv_v proc check_effective_target_riscv_v_ok { } { -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: Re：[PATCH v4] RISC-V: Handle differences between XTheadvector and Vector 2024-01-02 9:48 ` joshua @ 2024-01-02 9:52 ` juzhe.zhong 0 siblings, 0 replies; 133+ messages in thread From: juzhe.zhong @ 2024-01-02 9:52 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 72425 bytes --] There are 3 patterns in vector.md are VLS modes. I saw theadvector disables VLSmodes: bool vls_mode_valid_p (machine_mode vls_mode) { - if (!TARGET_VECTOR) + if (!TARGET_VECTOR || TARGET_XTHEADVECTOR) return false; So it should be changing only 2 patterns in vector.md juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-02 17:48 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v4] RISC-V: Handle differences between XTheadvector and Vector + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (mode, dest, src, + RVV_VLMAX, GEN_INT(VLMAX))); + return true; + } Move it outside legitimize_move It should be it: if (TARGET_THEADVECTOR) emit_th_move... DONE; else if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; If we move emit_insn (gen_pred_th_whole_mov...) outside legitimize, we need to modify the patterns which will call legitimize_move (). I have tried and that will result in 5 extra changes in vector.md. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月2日(星期二) 10:00 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v4] RISC-V: Handle differences between XTheadvector and Vector + if (TARGET_XTHEADVECTOR) + return false; Move it to : if (TARGET_VECTOR && stringop_strategy & STRATEGY_VECTOR) { bool ok = riscv_vector::expand_block_move (dest, src, length); if (ok) return true; } (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (match_test "!TARGET_XTHEADVECTOR || rtx_equal_p (op, const0_rtx)") + (match_operand 0 "const_csr_operand")))) It's hard to trace. Change it into : (ior 1. TARGET_THEADVECTOR && rtx_equal_p (op, const0_rtx) 2. !TAGEET_THEADVECTOR && const_csr_operand) + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (mode, dest, src, + RVV_VLMAX, GEN_INT(VLMAX))); + return true; + } Move it outside legitimize_move It should be it: if (TARGET_THEADVECTOR) emit_th_move... DONE; else if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; vsetvli issues: I wonder whether we can use ASM_OUTPUT_OPCODE to recognize "ta,ma"/"ta,mu"/"tu,ma"/"tu,mu" and replace these 4 variants by "". So that we don't have tail policy and mask policy in vsetvli ASM string. Another alternative approach is we can change vsetlvi ASM rule: "vset%i1vli\t%0,%1,e%2,%m3,t%p4,m%p5" if (TARGET_THEADVECTOR) ... else else if (code == CONST_INT) { /* Tail && Mask policy. */ asm_fprintf (file, "%s", IS_AGNOSTIC (UINTVAL (op)) ? "a" : "u"); } in riscv.cc. The benefit is that we can avoid adding all th_vsetvl patterns and invasive code changs in VSETVL PASS. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2023-12-29 12:21 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v4] RISC-V: Handle differences between XTheadvector and Vector This patch is to handle the differences in instruction generation between Vector and XTheadVector. In this version, we only support partial xtheadvector instructions that leverage directly from current RVV1.0 with simple adding "th." prefix. For different name xtheadvector instructions but share same patterns as RVV1.0 instructions, we will use ASM targethook to rewrite the whole string of the instructions in the following patches. For some vector patterns that cannot be avoided, we use "!TARGET_XTHEADVECTOR" to disable them in vector.md in order not to generate instructions that xtheadvector does not support, like vmv1r and vsext.vf2. gcc/ChangeLog: * config.gcc: Add files for XTheadVector intrinsics. * config/riscv/autovec.md: Guard XTheadVector. * config/riscv/riscv-string.cc (expand_block_move): Guard XTheadVector. * config/riscv/riscv-v.cc (legitimize_move): New expansion. (get_prefer_tail_policy): Give specific value for tail. (get_prefer_mask_policy): Give specific value for mask. (vls_mode_valid_p): Avoid autovec. * config/riscv/riscv-vector-builtins-shapes.cc (check_type): (build_one): New function. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): (DEF_THEAD_RVV_FUNCTION): Add new marcos. (check_required_extensions): (handle_pragma_vector): * config/riscv/riscv-vector-builtins.h (RVV_REQUIRE_VECTOR): (RVV_REQUIRE_XTHEADVECTOR): Add RVV_REQUIRE_VECTOR and RVV_REQUIRE_XTHEADVECTOR. (struct function_group_info): * config/riscv/riscv-vector-switch.def (ENTRY): Disable fractional mode for the XTheadVector extension. (TUPLE_ENTRY): Likewise. * config/riscv/riscv-vsetvl.cc: Add functions for xtheadvector. * config/riscv/riscv.cc (riscv_v_ext_vls_mode_p): Guard XTheadVector. (riscv_v_adjust_bytesize): Likewise. (riscv_preferred_simd_mode): Likewsie. (riscv_autovectorize_vector_modes): Likewise. (riscv_vector_mode_supported_any_target_p): Likewise. (TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P): Likewise. * config/riscv/vector-iterators.md: Remove fractional LMUL. * config/riscv/vector.md: Include thead-vector.md. * config/riscv/riscv_th_vector.h: New file. * config/riscv/thead-vector.md: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/base/pragma-1.c: Add XTheadVector. * gcc.target/riscv/rvv/base/abi-1.c: Exclude XTheadVector. * lib/target-supports.exp: Add target for XTheadVector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config.gcc | 2 +- gcc/config/riscv/autovec.md | 2 +- gcc/config/riscv/predicates.md | 3 +- gcc/config/riscv/riscv-string.cc | 3 + gcc/config/riscv/riscv-v.cc | 13 +- .../riscv/riscv-vector-builtins-bases.cc | 3 + .../riscv/riscv-vector-builtins-shapes.cc | 23 +++ gcc/config/riscv/riscv-vector-switch.def | 150 +++++++------- gcc/config/riscv/riscv-vsetvl.cc | 10 + gcc/config/riscv/riscv.cc | 20 +- gcc/config/riscv/riscv_th_vector.h | 49 +++++ gcc/config/riscv/thead-vector.md | 142 +++++++++++++ gcc/config/riscv/vector-iterators.md | 186 +++++++++--------- gcc/config/riscv/vector.md | 36 +++- .../gcc.target/riscv/rvv/base/abi-1.c | 2 +- .../gcc.target/riscv/rvv/base/pragma-1.c | 2 +- gcc/testsuite/lib/target-supports.exp | 12 ++ 17 files changed, 471 insertions(+), 187 deletions(-) create mode 100644 gcc/config/riscv/riscv_th_vector.h create mode 100644 gcc/config/riscv/thead-vector.md diff --git a/gcc/config.gcc b/gcc/config.gcc index f0676c830e8..1445d98c147 100644 --- a/gcc/config.gcc +++ b/gcc/config.gcc @@ -549,7 +549,7 @@ riscv*) extra_objs="${extra_objs} riscv-vector-builtins.o riscv-vector-builtins-shapes.o riscv-vector-builtins-bases.o" extra_objs="${extra_objs} thead.o riscv-target-attr.o" d_target_objs="riscv-d.o" - extra_headers="riscv_vector.h" + extra_headers="riscv_vector.h riscv_th_vector.h" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.cc" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.h" ;; diff --git a/gcc/config/riscv/autovec.md b/gcc/config/riscv/autovec.md index 8b8a92f10a1..1fac56c7095 100644 --- a/gcc/config/riscv/autovec.md +++ b/gcc/config/riscv/autovec.md @@ -2579,7 +2579,7 @@ [(match_operand 0 "register_operand") (match_operand 1 "memory_operand") (match_operand:ANYI 2 "const_int_operand")] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" { riscv_vector::expand_rawmemchr(<MODE>mode, operands[0], operands[1], operands[2]); diff --git a/gcc/config/riscv/predicates.md b/gcc/config/riscv/predicates.md index b87a6900841..c5ab86d5db9 100644 --- a/gcc/config/riscv/predicates.md +++ b/gcc/config/riscv/predicates.md @@ -428,7 +428,8 @@ ;; Predicates for the V extension. (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (match_test "!TARGET_XTHEADVECTOR || rtx_equal_p (op, const0_rtx)") + (match_operand 0 "const_csr_operand")))) (define_special_predicate "autovec_length_operand" (ior (match_operand 0 "pmode_register_operand") diff --git a/gcc/config/riscv/riscv-string.cc b/gcc/config/riscv/riscv-string.cc index 11c1f74d0b3..ec8f3486fd8 100644 --- a/gcc/config/riscv/riscv-string.cc +++ b/gcc/config/riscv/riscv-string.cc @@ -808,6 +808,9 @@ expand_block_move (rtx dst_in, rtx src_in, rtx length_in) bnez a2, loop # Any more? ret # Return */ + if (TARGET_XTHEADVECTOR) + return false; + gcc_assert (TARGET_VECTOR); HOST_WIDE_INT potential_ew diff --git a/gcc/config/riscv/riscv-v.cc b/gcc/config/riscv/riscv-v.cc index 038ab084a37..5e9e45aecd2 100644 --- a/gcc/config/riscv/riscv-v.cc +++ b/gcc/config/riscv/riscv-v.cc @@ -1523,6 +1523,13 @@ legitimize_move (rtx dest, rtx *srcp) return true; } + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (mode, dest, src, + RVV_VLMAX, GEN_INT(VLMAX))); + return true; + } + if (riscv_v_ext_vls_mode_p (mode)) { if (GET_MODE_NUNITS (mode).to_constant () <= 31) @@ -1772,7 +1779,7 @@ get_prefer_tail_policy () compiler pick up either agnostic or undisturbed. Maybe we will have a compile option like -mprefer=agnostic to set this value???. */ - return TAIL_ANY; + return TARGET_XTHEADVECTOR ? TAIL_AGNOSTIC : TAIL_ANY; } /* Get prefer mask policy. */ @@ -1783,7 +1790,7 @@ get_prefer_mask_policy () compiler pick up either agnostic or undisturbed. Maybe we will have a compile option like -mprefer=agnostic to set this value???. */ - return MASK_ANY; + return TARGET_XTHEADVECTOR ? MASK_UNDISTURBED : MASK_ANY; } /* Get avl_type rtx. */ @@ -4383,7 +4390,7 @@ cmp_lmul_gt_one (machine_mode mode) bool vls_mode_valid_p (machine_mode vls_mode) { - if (!TARGET_VECTOR) + if (!TARGET_VECTOR || TARGET_XTHEADVECTOR) return false; if (riscv_autovec_preference == RVV_SCALABLE) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index c51affde353..2918c07ebf3 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -133,6 +133,9 @@ public: = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + if (TARGET_XTHEADVECTOR) + return e.generate_insn (code_for_th_vsetvl_no_side_effects (Pmode)); + /* TAIL_ANY. */ e.add_input_operand (Pmode, gen_int_mode (get_prefer_tail_policy (), Pmode)); diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 4a754e0228f..6b49404a1fa 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -33,6 +33,25 @@ namespace riscv_vector { +/* Check whether the RETURN_TYPE and ARGUMENT_TYPES are + valid for the function. */ + +static bool +check_type (tree return_type, vec<tree> &argument_types) +{ + tree arg; + unsigned i; + + if (!return_type) + return false; + + FOR_EACH_VEC_ELT (argument_types, i, arg) + if (!arg) + return false; + + return true; +} + /* Add one function instance for GROUP, using operand suffix at index OI, mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ static void @@ -49,6 +68,10 @@ build_one (function_builder &b, const function_group_info &group, group.ops_infos.types[vec_type_idx].index); b.allocate_argument_types (function_instance, argument_types); b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + b.add_overloaded_function (function_instance, *group.shape); b.add_unique_function (function_instance, (*group.shape), return_type, argument_types); diff --git a/gcc/config/riscv/riscv-vector-switch.def b/gcc/config/riscv/riscv-vector-switch.def index 5c9f9bcbc3e..f7a66b34bae 100644 --- a/gcc/config/riscv/riscv-vector-switch.def +++ b/gcc/config/riscv/riscv-vector-switch.def @@ -68,9 +68,9 @@ Encode the ratio of SEW/LMUL into the mask types. #endif /* Disable modes if TARGET_MIN_VLEN == 32. */ -ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) -ENTRY (RVVMF32BI, true, LMUL_F4, 32) -ENTRY (RVVMF16BI, true, LMUL_F2, 16) +ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F8, 64) +ENTRY (RVVMF32BI, true, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F4, 32) +ENTRY (RVVMF16BI, true, TARGET_XTHEADVECTOR ? LMUL_1 : LMUL_F2 , 16) ENTRY (RVVMF8BI, true, LMUL_1, 8) ENTRY (RVVMF4BI, true, LMUL_2, 4) ENTRY (RVVMF2BI, true, LMUL_4, 2) @@ -81,39 +81,39 @@ ENTRY (RVVM8QI, true, LMUL_8, 1) ENTRY (RVVM4QI, true, LMUL_4, 2) ENTRY (RVVM2QI, true, LMUL_2, 4) ENTRY (RVVM1QI, true, LMUL_1, 8) -ENTRY (RVVMF2QI, true, LMUL_F2, 16) -ENTRY (RVVMF4QI, true, LMUL_F4, 32) -ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) +ENTRY (RVVMF2QI, !TARGET_XTHEADVECTOR, LMUL_F2, 16) +ENTRY (RVVMF4QI, !TARGET_XTHEADVECTOR, LMUL_F4, 32) +ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F8, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8HI, true, LMUL_8, 2) ENTRY (RVVM4HI, true, LMUL_4, 4) ENTRY (RVVM2HI, true, LMUL_2, 8) ENTRY (RVVM1HI, true, LMUL_1, 16) -ENTRY (RVVMF2HI, true, LMUL_F2, 32) -ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HI, !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_16. */ ENTRY (RVVM8HF, TARGET_VECTOR_ELEN_FP_16, LMUL_8, 2) ENTRY (RVVM4HF, TARGET_VECTOR_ELEN_FP_16, LMUL_4, 4) ENTRY (RVVM2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_2, 8) ENTRY (RVVM1HF, TARGET_VECTOR_ELEN_FP_16, LMUL_1, 16) -ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_F2, 32) -ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8SI, true, LMUL_8, 4) ENTRY (RVVM4SI, true, LMUL_4, 8) ENTRY (RVVM2SI, true, LMUL_2, 16) ENTRY (RVVM1SI, true, LMUL_1, 32) -ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_32. */ ENTRY (RVVM8SF, TARGET_VECTOR_ELEN_FP_32, LMUL_8, 4) ENTRY (RVVM4SF, TARGET_VECTOR_ELEN_FP_32, LMUL_4, 8) ENTRY (RVVM2SF, TARGET_VECTOR_ELEN_FP_32, LMUL_2, 16) ENTRY (RVVM1SF, TARGET_VECTOR_ELEN_FP_32, LMUL_1, 32) -ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if !TARGET_VECTOR_ELEN_64. */ ENTRY (RVVM8DI, TARGET_VECTOR_ELEN_64, LMUL_8, 8) @@ -140,127 +140,127 @@ ENTRY (RVVM1DF, TARGET_VECTOR_ELEN_FP_64, LMUL_1, 64) #endif TUPLE_ENTRY (RVVM1x8QI, true, RVVM1QI, 8, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x8QI, true, RVVMF2QI, 8, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x8QI, true, RVVMF4QI, 8, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 8, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x8QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 8, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x8QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 8, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 8, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x7QI, true, RVVM1QI, 7, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x7QI, true, RVVMF2QI, 7, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x7QI, true, RVVMF4QI, 7, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 7, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x7QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 7, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x7QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 7, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 7, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x6QI, true, RVVM1QI, 6, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x6QI, true, RVVMF2QI, 6, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x6QI, true, RVVMF4QI, 6, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 6, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x6QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 6, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x6QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 6, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 6, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x5QI, true, RVVM1QI, 5, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x5QI, true, RVVMF2QI, 5, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x5QI, true, RVVMF4QI, 5, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 5, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x5QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 5, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x5QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 5, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 5, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x4QI, true, RVVM2QI, 4, LMUL_2, 4) TUPLE_ENTRY (RVVM1x4QI, true, RVVM1QI, 4, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x4QI, true, RVVMF2QI, 4, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x4QI, true, RVVMF4QI, 4, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 4, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x4QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 4, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x4QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 4, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 4, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x3QI, true, RVVM2QI, 3, LMUL_2, 4) TUPLE_ENTRY (RVVM1x3QI, true, RVVM1QI, 3, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x3QI, true, RVVMF2QI, 3, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x3QI, true, RVVMF4QI, 3, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 3, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x3QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 3, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x3QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 3, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 3, LMUL_F8, 64) TUPLE_ENTRY (RVVM4x2QI, true, RVVM4QI, 2, LMUL_4, 2) TUPLE_ENTRY (RVVM2x2QI, true, RVVM2QI, 2, LMUL_2, 4) TUPLE_ENTRY (RVVM1x2QI, true, RVVM1QI, 2, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x2QI, true, RVVMF2QI, 2, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x2QI, true, RVVMF4QI, 2, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 2, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x2QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 2, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x2QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 2, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 2, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x8HI, true, RVVM1HI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HI, true, RVVMF2HI, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HI, true, RVVM1HI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HI, true, RVVMF2HI, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HI, true, RVVM1HI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HI, true, RVVMF2HI, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HI, true, RVVM1HI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HI, true, RVVMF2HI, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HI, true, RVVM2HI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HI, true, RVVM1HI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HI, true, RVVMF2HI, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HI, true, RVVM2HI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HI, true, RVVM1HI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HI, true, RVVMF2HI, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HI, true, RVVM4HI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HI, true, RVVM2HI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HI, true, RVVM1HI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HI, true, RVVMF2HI, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM4HF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8SI, true, RVVM1SI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SI, true, RVVM1SI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SI, true, RVVM1SI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SI, true, RVVM1SI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SI, true, RVVM2SI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SI, true, RVVM1SI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SI, true, RVVM2SI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SI, true, RVVM1SI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SI, true, RVVM4SI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SI, true, RVVM2SI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SI, true, RVVM1SI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM4SF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 8, LMUL_1, 16) TUPLE_ENTRY (RVVM1x7DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 7, LMUL_1, 16) diff --git a/gcc/config/riscv/riscv-vsetvl.cc b/gcc/config/riscv/riscv-vsetvl.cc index eabaef80f89..c726253c107 100644 --- a/gcc/config/riscv/riscv-vsetvl.cc +++ b/gcc/config/riscv/riscv-vsetvl.cc @@ -1117,6 +1117,16 @@ public: avl = GEN_INT (0); rtx sew = gen_int_mode (get_sew (), Pmode); rtx vlmul = gen_int_mode (get_vlmul (), Pmode); + + if (TARGET_XTHEADVECTOR) { + if (change_vtype_only_p ()) + return gen_th_vsetvl_vtype_change_only (sew, vlmul); + else if (has_vl () && !ignore_vl) + return gen_th_vsetvl (Pmode, get_vl (), avl, sew, vlmul); + else + return gen_th_vsetvl_discard_result (Pmode, avl, sew, vlmul); + } + rtx ta = gen_int_mode (get_ta (), Pmode); rtx ma = gen_int_mode (get_ma (), Pmode); diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index ea1d59d9cf2..a80bf8d1a74 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -1406,6 +1406,9 @@ riscv_v_adjust_bytesize (machine_mode mode, int scale) { if (riscv_v_ext_vector_mode_p (mode)) { + if (TARGET_XTHEADVECTOR) + return BYTES_PER_RISCV_VECTOR; + poly_int64 nunits = GET_MODE_NUNITS (mode); poly_int64 mode_size = GET_MODE_SIZE (mode); @@ -9965,7 +9968,7 @@ riscv_use_divmod_expander (void) static machine_mode riscv_preferred_simd_mode (scalar_mode mode) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::preferred_simd_mode (mode); return word_mode; @@ -10316,7 +10319,7 @@ riscv_mode_priority (int, int n) unsigned int riscv_autovectorize_vector_modes (vector_modes *modes, bool all) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::autovectorize_vector_modes (modes, all); return default_autovectorize_vector_modes (modes, all); @@ -10499,6 +10502,16 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) return false; } +/* Implements target hook vector_mode_supported_any_target_p. */ + +static bool +riscv_vector_mode_supported_any_target_p (machine_mode mode) +{ + if (TARGET_XTHEADVECTOR) + return false; + return true; +} + /* Initialize the GCC target structure. */ #undef TARGET_ASM_ALIGNED_HI_OP #define TARGET_ASM_ALIGNED_HI_OP "\t.half\t" @@ -10842,6 +10855,9 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) #undef TARGET_PREFERRED_ELSE_VALUE #define TARGET_PREFERRED_ELSE_VALUE riscv_preferred_else_value +#undef TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P +#define TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P riscv_vector_mode_supported_any_target_p + struct gcc_target targetm = TARGET_INITIALIZER; #include "gt-riscv.h" diff --git a/gcc/config/riscv/riscv_th_vector.h b/gcc/config/riscv/riscv_th_vector.h new file mode 100644 index 00000000000..6f47e0c90a4 --- /dev/null +++ b/gcc/config/riscv/riscv_th_vector.h @@ -0,0 +1,49 @@ +/* RISC-V 'XTheadVector' Extension intrinsics include file. + Copyright (C) 2022-2023 Free Software Foundation, Inc. + + This file is part of GCC. + + GCC is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published + by the Free Software Foundation; either version 3, or (at your + option) any later version. + + GCC is distributed in the hope that it will be useful, but WITHOUT + ANY WARRANTY; without even the implied warranty of MERCHANTABILITY + or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public + License for more details. + + Under Section 7 of GPL version 3, you are granted additional + permissions described in the GCC Runtime Library Exception, version + 3.1, as published by the Free Software Foundation. + + You should have received a copy of the GNU General Public License and + a copy of the GCC Runtime Library Exception along with this program; + see the files COPYING3 and COPYING.RUNTIME respectively. If not, see + <" target="_blank">http://www.gnu.org/licenses/>. */ + +#ifndef __RISCV_TH_VECTOR_H +#define __RISCV_TH_VECTOR_H + +#include <stdint.h> +#include <stddef.h> + +#ifndef __riscv_xtheadvector +#error "XTheadVector intrinsics require the xtheadvector extension." +#else +#ifdef __cplusplus +extern "C" { +#endif + +/* NOTE: This implementation of riscv_th_vector.h is intentionally short. It does + not define the RVV types and intrinsic functions directly in C and C++ + code, but instead uses the following pragma to tell GCC to insert the + necessary type and function definitions itself. The net effect is the + same, and the file is a complete implementation of riscv_th_vector.h. */ +#pragma riscv intrinsic "vector" + +#ifdef __cplusplus +} +#endif // __cplusplus +#endif // __riscv_xtheadvector +#endif // __RISCV_TH_ECTOR_H diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md new file mode 100644 index 00000000000..f0aa196d93b --- /dev/null +++ b/gcc/config/riscv/thead-vector.md @@ -0,0 +1,142 @@ +(define_c_enum "unspec" [ + UNSPEC_TH_VWLDST +]) + +(define_mode_iterator V_VLS_VT [V VLS VT]) +(define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) + +(define_split + [(set (match_operand:V_VB_VLS_VT 0 "reg_or_mem_operand") + (match_operand:V_VB_VLS_VT 1 "reg_or_mem_operand"))] + "TARGET_XTHEADVECTOR" + [(const_int 0)] + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + }) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:V_VLS_VT 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:V_VLS_VT + [(match_operand:V_VLS_VT 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2")]) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:VB 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:VB + [(match_operand:VB 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2") + (set (attr "sew") (const_int 8)) + (set (attr "vlmul") (symbol_ref "riscv_vector::LMUL_1"))]) + +(define_insn "@th_vsetvl<mode>" + [(set (match_operand:P 0 "register_operand" "=r") + (unspec:P [(match_operand:P 1 "vector_length_operand" "rK") + (match_operand 2 "const_int_operand" "i") + (match_operand 3 "const_int_operand" "i")] UNSPEC_VSETVL)) + (set (reg:SI VL_REGNUM) + (unspec:SI [(match_dup 1) + (match_dup 2) + (match_dup 3)] UNSPEC_VSETVL)) + (set (reg:SI VTYPE_REGNUM) + (unspec:SI [(match_dup 2) + (match_dup 3)] UNSPEC_VSETVL))] + "TARGET_XTHEADVECTOR" + "vsetvli\t%0,%1,e%2,%m3" + [(set_attr "type" "vsetvl") + (set_attr "mode" "<MODE>") + (set (attr "sew") (symbol_ref "INTVAL (operands[2])")) + (set (attr "vlmul") (symbol_ref "INTVAL (operands[3])"))]) + +;; vsetvl zero,zero,vtype instruction. +;; This pattern has no side effects and does not set X0 register. +(define_insn "th_vsetvl_vtype_change_only" + [(set (reg:SI VTYPE_REGNUM) + (unspec:SI + [(match_operand 0 "const_int_operand" "i") + (match_operand 1 "const_int_operand" "i")] UNSPEC_VSETVL))] + "TARGET_XTHEADVECTOR" + "vsetvli\tzero,zero,e%0,%m1" + [(set_attr "type" "vsetvl") + (set_attr "mode" "SI") + (set (attr "sew") (symbol_ref "INTVAL (operands[0])")) + (set (attr "vlmul") (symbol_ref "INTVAL (operands[1])"))]) + +;; vsetvl zero,rs1,vtype instruction. +;; The reason we need this pattern since we should avoid setting X0 register +;; in vsetvl instruction pattern. +(define_insn "@th_vsetvl_discard_result<mode>" + [(set (reg:SI VL_REGNUM) + (unspec:SI [(match_operand:P 0 "vector_length_operand" "rK") + (match_operand 1 "const_int_operand" "i") + (match_operand 2 "const_int_operand" "i")] UNSPEC_VSETVL)) + (set (reg:SI VTYPE_REGNUM) + (unspec:SI [(match_dup 1) + (match_dup 2)] UNSPEC_VSETVL))] + "TARGET_XTHEADVECTOR" + "vsetvli\tzero,%0,e%1,%m2" + [(set_attr "type" "vsetvl") + (set_attr "mode" "<MODE>") + (set (attr "sew") (symbol_ref "INTVAL (operands[1])")) + (set (attr "vlmul") (symbol_ref "INTVAL (operands[2])"))]) + +;; It's emit by vsetvl/vsetvlmax intrinsics with no side effects. +;; Since we have many optmization passes from "expand" to "reload_completed", +;; such pattern can allow us gain benefits of these optimizations. +(define_insn_and_split "@th_vsetvl<mode>_no_side_effects" + [(set (match_operand:P 0 "register_operand" "=r") + (unspec:P [(match_operand:P 1 "vector_length_operand" "rK") + (match_operand 2 "const_int_operand" "i") + (match_operand 3 "const_int_operand" "i")] UNSPEC_VSETVL))] + "TARGET_XTHEADVECTOR" + "#" + "&& epilogue_completed" + [(parallel + [(set (match_dup 0) + (unspec:P [(match_dup 1) (match_dup 2) (match_dup 3)] UNSPEC_VSETVL)) + (set (reg:SI VL_REGNUM) + (unspec:SI [(match_dup 1) (match_dup 2) (match_dup 3)] UNSPEC_VSETVL)) + (set (reg:SI VTYPE_REGNUM) + (unspec:SI [(match_dup 2) (match_dup 3)] UNSPEC_VSETVL))])] + "" + [(set_attr "type" "vsetvl") + (set_attr "mode" "SI")]) diff --git a/gcc/config/riscv/vector-iterators.md b/gcc/config/riscv/vector-iterators.md index 5f5f7b5b986..c0fc7a2441d 100644 --- a/gcc/config/riscv/vector-iterators.md +++ b/gcc/config/riscv/vector-iterators.md @@ -109,11 +109,11 @@ ]) (define_mode_iterator VI [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -128,11 +128,11 @@ ;; allow the instruction and mode to be matched during combine et al. (define_mode_iterator VF [ (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -140,11 +140,11 @@ (define_mode_iterator VF_ZVFHMIN [ (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -271,16 +271,16 @@ ]) (define_mode_iterator VEEWEXT2 [ - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -290,10 +290,10 @@ ]) (define_mode_iterator VEEWEXT4 [ - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -311,59 +311,59 @@ ]) (define_mode_iterator VEEWTRUNC2 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") (RVVM4SI "TARGET_64BIT") (RVVM2SI "TARGET_64BIT") (RVVM1SI "TARGET_64BIT") - (RVVMF2SI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEEWTRUNC4 [ - RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM2HI "TARGET_64BIT") (RVVM1HI "TARGET_64BIT") - (RVVMF2HI "TARGET_64BIT") - (RVVMF4HI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2HI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") (RVVM1HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEEWTRUNC8 [ (RVVM1QI "TARGET_64BIT") - (RVVMF2QI "TARGET_64BIT") - (RVVMF4QI "TARGET_64BIT") - (RVVMF8QI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2QI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF4QI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEI16 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -452,11 +452,11 @@ ]) (define_mode_iterator VFULLI [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_FULL_V") (RVVM4DI "TARGET_FULL_V") (RVVM2DI "TARGET_FULL_V") (RVVM1DI "TARGET_FULL_V") @@ -509,17 +509,17 @@ ]) (define_mode_iterator VI_QH [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") ]) (define_mode_iterator VI_QHS [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -560,11 +560,11 @@ ]) (define_mode_iterator VI_QHS_NO_M8 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -603,11 +603,11 @@ (define_mode_iterator VF_HS [ (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (V1HF "riscv_vector::vls_mode_valid_p (V1HFmode) && TARGET_ZVFH") (V2HF "riscv_vector::vls_mode_valid_p (V2HFmode) && TARGET_ZVFH") @@ -638,12 +638,12 @@ (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") - (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (V1HF "riscv_vector::vls_mode_valid_p (V1HFmode) && TARGET_ZVFH") (V2HF "riscv_vector::vls_mode_valid_p (V2HFmode) && TARGET_ZVFH") @@ -674,11 +674,11 @@ ]) (define_mode_iterator V_VLSI_QHS [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -756,27 +756,27 @@ ;; E.g. when index mode = RVVM8QImde and Pmode = SImode, if it is not zero_extend or ;; scalar != 1, such gather/scatter is not allowed since we don't have RVVM32SImode. (define_mode_iterator RATIO64 [ - (RVVMF8QI "TARGET_MIN_VLEN > 32") - (RVVMF4HI "TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM1DI "TARGET_VECTOR_ELEN_64") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") ]) (define_mode_iterator RATIO32 [ - RVVMF4QI - RVVMF2HI + (RVVMF4QI "!TARGET_XTHEADVECTOR") + (RVVMF2HI "!TARGET_XTHEADVECTOR") RVVM1SI (RVVM2DI "TARGET_VECTOR_ELEN_64") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") ]) (define_mode_iterator RATIO16 [ - RVVMF2QI + (RVVMF2QI "!TARGET_XTHEADVECTOR") RVVM1HI RVVM2SI (RVVM4DI "TARGET_VECTOR_ELEN_64") @@ -814,21 +814,21 @@ ]) (define_mode_iterator RATIO64I [ - (RVVMF8QI "TARGET_MIN_VLEN > 32") - (RVVMF4HI "TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM1DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") ]) (define_mode_iterator RATIO32I [ - RVVMF4QI - RVVMF2HI + (RVVMF4QI "!TARGET_XTHEADVECTOR") + (RVVMF2HI "!TARGET_XTHEADVECTOR") RVVM1SI (RVVM2DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") ]) (define_mode_iterator RATIO16I [ - RVVMF2QI + (RVVMF2QI "!TARGET_XTHEADVECTOR") RVVM1HI RVVM2SI (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") @@ -873,21 +873,21 @@ ]) (define_mode_iterator V_FRACT [ - RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") ]) (define_mode_iterator VWEXTI [ - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -933,7 +933,7 @@ (RVVM4SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -966,7 +966,7 @@ (RVVM4SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -996,7 +996,7 @@ (define_mode_iterator VWCONVERTI [ (RVVM8SI "TARGET_ZVFH") (RVVM4SI "TARGET_ZVFH") (RVVM2SI "TARGET_ZVFH") (RVVM1SI "TARGET_ZVFH") - (RVVMF2SI "TARGET_ZVFH") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_ZVFH") (RVVM8DI "TARGET_VECTOR_ELEN_64 && TARGET_VECTOR_ELEN_FP_32") (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_VECTOR_ELEN_FP_32") @@ -1045,7 +1045,7 @@ ]) (define_mode_iterator VQEXTI [ - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -1456,11 +1456,11 @@ ;; VINDEXED [VI8 VI16 VI32 (VI64 "TARGET_64BIT")]. (define_mode_iterator VINDEXED [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") @@ -1468,12 +1468,12 @@ (RVVM1DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64 && TARGET_64BIT") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64 && TARGET_64BIT") @@ -3173,11 +3173,11 @@ (define_mode_iterator V_VLS_F_CONVERT_SI [ (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") - (RVVMF2HF "TARGET_ZVFH") (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") @@ -3290,12 +3290,12 @@ ]) (define_mode_iterator V_VLS_F_CONVERT_DI [ - (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index b5a9055cdc4..9e7a487759a 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -83,7 +83,7 @@ ;; check. However, we need default value of SEW for vsetvl instruction since there ;; is no field for ratio in the vsetvl instruction encoding. (define_attr "sew" "" - (cond [(eq_attr "mode" "RVVMF64BI,RVVMF32BI,RVVMF16BI,RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ + (cond [(eq_attr "mode" "RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ RVVM8QI,RVVM4QI,RVVM2QI,RVVM1QI,RVVMF2QI,RVVMF4QI,RVVMF8QI,\ RVVM1x8QI,RVVMF2x8QI,RVVMF4x8QI,RVVMF8x8QI,\ RVVM1x7QI,RVVMF2x7QI,RVVMF4x7QI,RVVMF8x7QI,\ @@ -95,6 +95,18 @@ V1QI,V2QI,V4QI,V8QI,V16QI,V32QI,V64QI,V128QI,V256QI,V512QI,V1024QI,V2048QI,V4096QI,\ V1BI,V2BI,V4BI,V8BI,V16BI,V32BI,V64BI,V128BI,V256BI,V512BI,V1024BI,V2048BI,V4096BI") (const_int 8) + (eq_attr "mode" "RVVMF16BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 16) + (const_int 8)) + (eq_attr "mode" "RVVMF32BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 32) + (const_int 8)) + (eq_attr "mode" "RVVMF64BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 64) + (const_int 8)) (eq_attr "mode" "RVVM8HI,RVVM4HI,RVVM2HI,RVVM1HI,RVVMF2HI,RVVMF4HI,\ RVVM1x8HI,RVVMF2x8HI,RVVMF4x8HI,\ RVVM1x7HI,RVVMF2x7HI,RVVMF4x7HI,\ @@ -155,9 +167,9 @@ (eq_attr "mode" "RVVM4QI,RVVMF2BI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2QI,RVVMF4BI") (symbol_ref "riscv_vector::LMUL_2") (eq_attr "mode" "RVVM1QI,RVVMF8BI") (symbol_ref "riscv_vector::LMUL_1") - (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "riscv_vector::LMUL_F2") - (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "riscv_vector::LMUL_F4") - (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "riscv_vector::LMUL_F8") + (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F2") + (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F4") + (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F8") (eq_attr "mode" "RVVM8HI") (symbol_ref "riscv_vector::LMUL_8") (eq_attr "mode" "RVVM4HI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2HI") (symbol_ref "riscv_vector::LMUL_2") @@ -428,6 +440,10 @@ vislide1up,vislide1down,vfslide1up,vfslide1down,\ vgather,vcompress,vlsegdux,vlsegdox,vssegtux,vssegtox") (const_int INVALID_ATTRIBUTE) + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE) (eq_attr "mode" "RVVM8QI,RVVM1BI") (const_int 1) (eq_attr "mode" "RVVM4QI,RVVMF2BI") (const_int 2) (eq_attr "mode" "RVVM2QI,RVVMF4BI") (const_int 4) @@ -888,6 +904,8 @@ (symbol_ref "riscv_vector::FRM_DYN")] (symbol_ref "riscv_vector::FRM_NONE"))) +(include "thead-vector.md") + ;; ----------------------------------------------------------------- ;; ---- Miscellaneous Operations ;; ----------------------------------------------------------------- @@ -1097,7 +1115,7 @@ (define_insn "*mov<mode>_whole" [(set (match_operand:V_WHOLE 0 "reg_or_mem_operand" "=vr, m,vr") (match_operand:V_WHOLE 1 "reg_or_mem_operand" " m,vr,vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "@ vl%m1re<sew>.v\t%0,%1 vs%m1r.v\t%1,%0 @@ -1125,7 +1143,7 @@ (define_insn "*mov<mode>" [(set (match_operand:VB 0 "register_operand" "=vr") (match_operand:VB 1 "register_operand" " vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "vmv1r.v\t%0,%1" [(set_attr "type" "vmov") (set_attr "mode" "<MODE>")]) @@ -3680,7 +3698,7 @@ (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3701,7 +3719,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3722,7 +3740,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c index 2e0e12aa045..2eef9e1e1a8 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c @@ -1,4 +1,4 @@ -/* { dg-do compile } */ +/* { dg-do compile { target { ! riscv_xtheadvector } } } */ /* { dg-skip-if "test rvv intrinsic" { *-*-* } { "*" } { "-march=rv*v*" } } */ void foo0 () {__rvv_bool64_t t;} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c index 3d81b179235..ef329e30785 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c @@ -1,4 +1,4 @@ /* { dg-do compile } */ /* { dg-options "-O3 -march=rv32gc -mabi=ilp32d" } */ -#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' extension enabled} } */ +#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' or 'XTHEADVECTOR' extension enabled} } */ diff --git a/gcc/testsuite/lib/target-supports.exp b/gcc/testsuite/lib/target-supports.exp index 7f13ff0ca56..70df6b1401c 100644 --- a/gcc/testsuite/lib/target-supports.exp +++ b/gcc/testsuite/lib/target-supports.exp @@ -1952,6 +1952,18 @@ proc check_effective_target_riscv_zbb { } { }] } +# Return 1 if the target arch supports the XTheadVector extension, 0 otherwise. +# Cache the result. + +proc check_effective_target_riscv_xtheadvector { } { + return [check_no_compiler_messages riscv_ext_xtheadvector assembly { + #ifndef __riscv_xtheadvector + #error "Not __riscv_xtheadvector" + #endif + }] +} + + # Return 1 if we can execute code when using dg-add-options riscv_v proc check_effective_target_riscv_v_ok { } { -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* [PATCH v4] RISC-V: Handle differences between XTheadvector and Vector 2023-12-29 4:21 ` [PATCH v4] RISC-V: Handle differences between XTheadvector and Vector Jun Sha (Joshua) 2024-01-02 2:00 ` juzhe.zhong @ 2024-01-02 12:39 ` Jun Sha (Joshua) 2024-01-03 3:11 ` Kito Cheng 2024-01-03 6:15 ` Jun Sha (Joshua) 1 sibling, 2 replies; 133+ messages in thread From: Jun Sha (Joshua) @ 2024-01-02 12:39 UTC (permalink / raw) To: gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, Jun Sha (Joshua), Jin Ma, Xianmiao Qu This patch is to handle the differences in instruction generation between Vector and XTheadVector. In this version, we only support partial xtheadvector instructions that leverage directly from current RVV1.0 with simple adding "th." prefix. For different name xtheadvector instructions but share same patterns as RVV1.0 instructions, we will use ASM targethook to rewrite the whole string of the instructions in the following patches. For some vector patterns that cannot be avoided, we use "!TARGET_XTHEADVECTOR" to disable them in vector.md in order not to generate instructions that xtheadvector does not support, like vmv1r and vsext.vf2. gcc/ChangeLog: * config.gcc: Add files for XTheadVector intrinsics. * config/riscv/autovec.md: Guard XTheadVector. * config/riscv/riscv-string.cc (expand_block_move): Guard XTheadVector. (get_prefer_tail_policy): Give specific value for tail. (get_prefer_mask_policy): Give specific value for mask. (vls_mode_valid_p): Avoid autovec. * config/riscv/riscv-vector-builtins-shapes.cc (check_type): (build_one): New function. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): (DEF_THEAD_RVV_FUNCTION): Add new marcos. (check_required_extensions): (handle_pragma_vector): * config/riscv/riscv-vector-builtins.h (RVV_REQUIRE_VECTOR): (RVV_REQUIRE_XTHEADVECTOR): Add RVV_REQUIRE_VECTOR and RVV_REQUIRE_XTHEADVECTOR. (struct function_group_info): * config/riscv/riscv-vector-switch.def (ENTRY): Disable fractional mode for the XTheadVector extension. (TUPLE_ENTRY): Likewise. * config/riscv/riscv-vsetvl.cc: Add functions for xtheadvector. * config/riscv/riscv.cc (riscv_v_ext_vls_mode_p): Guard XTheadVector. (riscv_v_adjust_bytesize): Likewise. (riscv_preferred_simd_mode): Likewsie. (riscv_autovectorize_vector_modes): Likewise. (riscv_vector_mode_supported_any_target_p): Likewise. (TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P): Likewise. * config/riscv/vector-iterators.md: Remove fractional LMUL. * config/riscv/vector.md: Include thead-vector.md. * config/riscv/riscv_th_vector.h: New file. * config/riscv/thead-vector.md: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/base/pragma-1.c: Add XTheadVector. * gcc.target/riscv/rvv/base/abi-1.c: Exclude XTheadVector. * lib/target-supports.exp: Add target for XTheadVector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config.gcc | 2 +- gcc/config/riscv/autovec.md | 2 +- gcc/config/riscv/predicates.md | 4 +- gcc/config/riscv/riscv-string.cc | 3 +- gcc/config/riscv/riscv-v.cc | 6 +- .../riscv/riscv-vector-builtins-bases.cc | 48 +++-- .../riscv/riscv-vector-builtins-shapes.cc | 23 +++ gcc/config/riscv/riscv-vector-switch.def | 150 +++++++------- gcc/config/riscv/riscv.cc | 20 +- gcc/config/riscv/riscv_th_vector.h | 49 +++++ gcc/config/riscv/thead-vector.md | 69 +++++++ gcc/config/riscv/vector-iterators.md | 186 +++++++++--------- gcc/config/riscv/vector.md | 55 ++++-- .../gcc.target/riscv/rvv/base/abi-1.c | 2 +- .../gcc.target/riscv/rvv/base/pragma-1.c | 2 +- gcc/testsuite/lib/target-supports.exp | 12 ++ 16 files changed, 425 insertions(+), 208 deletions(-) create mode 100644 gcc/config/riscv/riscv_th_vector.h create mode 100644 gcc/config/riscv/thead-vector.md diff --git a/gcc/config.gcc b/gcc/config.gcc index f0676c830e8..1445d98c147 100644 --- a/gcc/config.gcc +++ b/gcc/config.gcc @@ -549,7 +549,7 @@ riscv*) extra_objs="${extra_objs} riscv-vector-builtins.o riscv-vector-builtins-shapes.o riscv-vector-builtins-bases.o" extra_objs="${extra_objs} thead.o riscv-target-attr.o" d_target_objs="riscv-d.o" - extra_headers="riscv_vector.h" + extra_headers="riscv_vector.h riscv_th_vector.h" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.cc" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.h" ;; diff --git a/gcc/config/riscv/autovec.md b/gcc/config/riscv/autovec.md index 8b8a92f10a1..1fac56c7095 100644 --- a/gcc/config/riscv/autovec.md +++ b/gcc/config/riscv/autovec.md @@ -2579,7 +2579,7 @@ [(match_operand 0 "register_operand") (match_operand 1 "memory_operand") (match_operand:ANYI 2 "const_int_operand")] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" { riscv_vector::expand_rawmemchr(<MODE>mode, operands[0], operands[1], operands[2]); diff --git a/gcc/config/riscv/predicates.md b/gcc/config/riscv/predicates.md index b87a6900841..6e36c5afae1 100644 --- a/gcc/config/riscv/predicates.md +++ b/gcc/config/riscv/predicates.md @@ -428,7 +428,9 @@ ;; Predicates for the V extension. (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (ior (match_test "TARGET_XTHEADVECTOR && rtx_equal_p (op, const0_rtx)") + (match_test "!TARGET_XTHEADVECTOR")) + (match_operand 0 "const_csr_operand")))) (define_special_predicate "autovec_length_operand" (ior (match_operand 0 "pmode_register_operand") diff --git a/gcc/config/riscv/riscv-string.cc b/gcc/config/riscv/riscv-string.cc index 11c1f74d0b3..f431b562222 100644 --- a/gcc/config/riscv/riscv-string.cc +++ b/gcc/config/riscv/riscv-string.cc @@ -773,7 +773,8 @@ riscv_expand_block_move_scalar (rtx dest, rtx src, rtx length) bool riscv_expand_block_move (rtx dest, rtx src, rtx length) { - if (TARGET_VECTOR && stringop_strategy & STRATEGY_VECTOR) + if ((TARGET_VECTOR && !TARGET_XTHEADVECTOR) + && stringop_strategy & STRATEGY_VECTOR) { bool ok = riscv_vector::expand_block_move (dest, src, length); if (ok) diff --git a/gcc/config/riscv/riscv-v.cc b/gcc/config/riscv/riscv-v.cc index 038ab084a37..eb061b593a5 100644 --- a/gcc/config/riscv/riscv-v.cc +++ b/gcc/config/riscv/riscv-v.cc @@ -1772,7 +1772,7 @@ get_prefer_tail_policy () compiler pick up either agnostic or undisturbed. Maybe we will have a compile option like -mprefer=agnostic to set this value???. */ - return TAIL_ANY; + return TARGET_XTHEADVECTOR ? TAIL_AGNOSTIC : TAIL_ANY; } /* Get prefer mask policy. */ @@ -1783,7 +1783,7 @@ get_prefer_mask_policy () compiler pick up either agnostic or undisturbed. Maybe we will have a compile option like -mprefer=agnostic to set this value???. */ - return MASK_ANY; + return TARGET_XTHEADVECTOR ? MASK_UNDISTURBED : MASK_ANY; } /* Get avl_type rtx. */ @@ -4383,7 +4383,7 @@ cmp_lmul_gt_one (machine_mode mode) bool vls_mode_valid_p (machine_mode vls_mode) { - if (!TARGET_VECTOR) + if (!TARGET_VECTOR || TARGET_XTHEADVECTOR) return false; if (riscv_autovec_preference == RVV_SCALABLE) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index c51affde353..22d877af9be 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -115,23 +115,37 @@ public: tree type = builtin_types[e.type.index].vector; machine_mode mode = TYPE_MODE (type); - /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. - - - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 - - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 - - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 - - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 - - e8,m2/e16,m4/e32,m8 --> e8m2 - - e8,m4/e16,m8 --> e8m4 - - e8,m8 --> e8m8 - */ - /* SEW. */ - e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); - - /* LMUL. */ - machine_mode e8_mode - = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); - e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + + if (TARGET_XTHEADVECTOR) + { + machine_mode inner_mode = GET_MODE_INNER (mode); + /* SEW. */ + e.add_input_operand (Pmode, + gen_int_mode (GET_MODE_BITSIZE (inner_mode), Pmode)); + /* LMUL. */ + e.add_input_operand (Pmode, + gen_int_mode (get_vlmul (mode), Pmode)); + } + else + { + /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. + + - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 + - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 + - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 + - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 + - e8,m2/e16,m4/e32,m8 --> e8m2 + - e8,m4/e16,m8 --> e8m4 + - e8,m8 --> e8m8 + */ + /* SEW. */ + e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); + + /* LMUL. */ + machine_mode e8_mode + = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); + e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + } /* TAIL_ANY. */ e.add_input_operand (Pmode, diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 4a754e0228f..6b49404a1fa 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -33,6 +33,25 @@ namespace riscv_vector { +/* Check whether the RETURN_TYPE and ARGUMENT_TYPES are + valid for the function. */ + +static bool +check_type (tree return_type, vec<tree> &argument_types) +{ + tree arg; + unsigned i; + + if (!return_type) + return false; + + FOR_EACH_VEC_ELT (argument_types, i, arg) + if (!arg) + return false; + + return true; +} + /* Add one function instance for GROUP, using operand suffix at index OI, mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ static void @@ -49,6 +68,10 @@ build_one (function_builder &b, const function_group_info &group, group.ops_infos.types[vec_type_idx].index); b.allocate_argument_types (function_instance, argument_types); b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + b.add_overloaded_function (function_instance, *group.shape); b.add_unique_function (function_instance, (*group.shape), return_type, argument_types); diff --git a/gcc/config/riscv/riscv-vector-switch.def b/gcc/config/riscv/riscv-vector-switch.def index 5c9f9bcbc3e..f7a66b34bae 100644 --- a/gcc/config/riscv/riscv-vector-switch.def +++ b/gcc/config/riscv/riscv-vector-switch.def @@ -68,9 +68,9 @@ Encode the ratio of SEW/LMUL into the mask types. #endif /* Disable modes if TARGET_MIN_VLEN == 32. */ -ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) -ENTRY (RVVMF32BI, true, LMUL_F4, 32) -ENTRY (RVVMF16BI, true, LMUL_F2, 16) +ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F8, 64) +ENTRY (RVVMF32BI, true, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F4, 32) +ENTRY (RVVMF16BI, true, TARGET_XTHEADVECTOR ? LMUL_1 : LMUL_F2 , 16) ENTRY (RVVMF8BI, true, LMUL_1, 8) ENTRY (RVVMF4BI, true, LMUL_2, 4) ENTRY (RVVMF2BI, true, LMUL_4, 2) @@ -81,39 +81,39 @@ ENTRY (RVVM8QI, true, LMUL_8, 1) ENTRY (RVVM4QI, true, LMUL_4, 2) ENTRY (RVVM2QI, true, LMUL_2, 4) ENTRY (RVVM1QI, true, LMUL_1, 8) -ENTRY (RVVMF2QI, true, LMUL_F2, 16) -ENTRY (RVVMF4QI, true, LMUL_F4, 32) -ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) +ENTRY (RVVMF2QI, !TARGET_XTHEADVECTOR, LMUL_F2, 16) +ENTRY (RVVMF4QI, !TARGET_XTHEADVECTOR, LMUL_F4, 32) +ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F8, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8HI, true, LMUL_8, 2) ENTRY (RVVM4HI, true, LMUL_4, 4) ENTRY (RVVM2HI, true, LMUL_2, 8) ENTRY (RVVM1HI, true, LMUL_1, 16) -ENTRY (RVVMF2HI, true, LMUL_F2, 32) -ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HI, !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_16. */ ENTRY (RVVM8HF, TARGET_VECTOR_ELEN_FP_16, LMUL_8, 2) ENTRY (RVVM4HF, TARGET_VECTOR_ELEN_FP_16, LMUL_4, 4) ENTRY (RVVM2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_2, 8) ENTRY (RVVM1HF, TARGET_VECTOR_ELEN_FP_16, LMUL_1, 16) -ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_F2, 32) -ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8SI, true, LMUL_8, 4) ENTRY (RVVM4SI, true, LMUL_4, 8) ENTRY (RVVM2SI, true, LMUL_2, 16) ENTRY (RVVM1SI, true, LMUL_1, 32) -ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_32. */ ENTRY (RVVM8SF, TARGET_VECTOR_ELEN_FP_32, LMUL_8, 4) ENTRY (RVVM4SF, TARGET_VECTOR_ELEN_FP_32, LMUL_4, 8) ENTRY (RVVM2SF, TARGET_VECTOR_ELEN_FP_32, LMUL_2, 16) ENTRY (RVVM1SF, TARGET_VECTOR_ELEN_FP_32, LMUL_1, 32) -ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if !TARGET_VECTOR_ELEN_64. */ ENTRY (RVVM8DI, TARGET_VECTOR_ELEN_64, LMUL_8, 8) @@ -140,127 +140,127 @@ ENTRY (RVVM1DF, TARGET_VECTOR_ELEN_FP_64, LMUL_1, 64) #endif TUPLE_ENTRY (RVVM1x8QI, true, RVVM1QI, 8, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x8QI, true, RVVMF2QI, 8, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x8QI, true, RVVMF4QI, 8, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 8, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x8QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 8, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x8QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 8, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 8, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x7QI, true, RVVM1QI, 7, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x7QI, true, RVVMF2QI, 7, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x7QI, true, RVVMF4QI, 7, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 7, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x7QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 7, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x7QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 7, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 7, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x6QI, true, RVVM1QI, 6, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x6QI, true, RVVMF2QI, 6, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x6QI, true, RVVMF4QI, 6, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 6, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x6QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 6, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x6QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 6, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 6, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x5QI, true, RVVM1QI, 5, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x5QI, true, RVVMF2QI, 5, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x5QI, true, RVVMF4QI, 5, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 5, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x5QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 5, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x5QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 5, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 5, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x4QI, true, RVVM2QI, 4, LMUL_2, 4) TUPLE_ENTRY (RVVM1x4QI, true, RVVM1QI, 4, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x4QI, true, RVVMF2QI, 4, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x4QI, true, RVVMF4QI, 4, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 4, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x4QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 4, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x4QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 4, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 4, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x3QI, true, RVVM2QI, 3, LMUL_2, 4) TUPLE_ENTRY (RVVM1x3QI, true, RVVM1QI, 3, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x3QI, true, RVVMF2QI, 3, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x3QI, true, RVVMF4QI, 3, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 3, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x3QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 3, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x3QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 3, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 3, LMUL_F8, 64) TUPLE_ENTRY (RVVM4x2QI, true, RVVM4QI, 2, LMUL_4, 2) TUPLE_ENTRY (RVVM2x2QI, true, RVVM2QI, 2, LMUL_2, 4) TUPLE_ENTRY (RVVM1x2QI, true, RVVM1QI, 2, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x2QI, true, RVVMF2QI, 2, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x2QI, true, RVVMF4QI, 2, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 2, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x2QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 2, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x2QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 2, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 2, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x8HI, true, RVVM1HI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HI, true, RVVMF2HI, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HI, true, RVVM1HI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HI, true, RVVMF2HI, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HI, true, RVVM1HI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HI, true, RVVMF2HI, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HI, true, RVVM1HI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HI, true, RVVMF2HI, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HI, true, RVVM2HI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HI, true, RVVM1HI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HI, true, RVVMF2HI, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HI, true, RVVM2HI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HI, true, RVVM1HI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HI, true, RVVMF2HI, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HI, true, RVVM4HI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HI, true, RVVM2HI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HI, true, RVVM1HI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HI, true, RVVMF2HI, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM4HF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8SI, true, RVVM1SI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SI, true, RVVM1SI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SI, true, RVVM1SI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SI, true, RVVM1SI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SI, true, RVVM2SI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SI, true, RVVM1SI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SI, true, RVVM2SI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SI, true, RVVM1SI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SI, true, RVVM4SI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SI, true, RVVM2SI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SI, true, RVVM1SI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM4SF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 8, LMUL_1, 16) TUPLE_ENTRY (RVVM1x7DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 7, LMUL_1, 16) diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index ea1d59d9cf2..a80bf8d1a74 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -1406,6 +1406,9 @@ riscv_v_adjust_bytesize (machine_mode mode, int scale) { if (riscv_v_ext_vector_mode_p (mode)) { + if (TARGET_XTHEADVECTOR) + return BYTES_PER_RISCV_VECTOR; + poly_int64 nunits = GET_MODE_NUNITS (mode); poly_int64 mode_size = GET_MODE_SIZE (mode); @@ -9965,7 +9968,7 @@ riscv_use_divmod_expander (void) static machine_mode riscv_preferred_simd_mode (scalar_mode mode) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::preferred_simd_mode (mode); return word_mode; @@ -10316,7 +10319,7 @@ riscv_mode_priority (int, int n) unsigned int riscv_autovectorize_vector_modes (vector_modes *modes, bool all) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::autovectorize_vector_modes (modes, all); return default_autovectorize_vector_modes (modes, all); @@ -10499,6 +10502,16 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) return false; } +/* Implements target hook vector_mode_supported_any_target_p. */ + +static bool +riscv_vector_mode_supported_any_target_p (machine_mode mode) +{ + if (TARGET_XTHEADVECTOR) + return false; + return true; +} + /* Initialize the GCC target structure. */ #undef TARGET_ASM_ALIGNED_HI_OP #define TARGET_ASM_ALIGNED_HI_OP "\t.half\t" @@ -10842,6 +10855,9 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) #undef TARGET_PREFERRED_ELSE_VALUE #define TARGET_PREFERRED_ELSE_VALUE riscv_preferred_else_value +#undef TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P +#define TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P riscv_vector_mode_supported_any_target_p + struct gcc_target targetm = TARGET_INITIALIZER; #include "gt-riscv.h" diff --git a/gcc/config/riscv/riscv_th_vector.h b/gcc/config/riscv/riscv_th_vector.h new file mode 100644 index 00000000000..6f47e0c90a4 --- /dev/null +++ b/gcc/config/riscv/riscv_th_vector.h @@ -0,0 +1,49 @@ +/* RISC-V 'XTheadVector' Extension intrinsics include file. + Copyright (C) 2022-2023 Free Software Foundation, Inc. + + This file is part of GCC. + + GCC is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published + by the Free Software Foundation; either version 3, or (at your + option) any later version. + + GCC is distributed in the hope that it will be useful, but WITHOUT + ANY WARRANTY; without even the implied warranty of MERCHANTABILITY + or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public + License for more details. + + Under Section 7 of GPL version 3, you are granted additional + permissions described in the GCC Runtime Library Exception, version + 3.1, as published by the Free Software Foundation. + + You should have received a copy of the GNU General Public License and + a copy of the GCC Runtime Library Exception along with this program; + see the files COPYING3 and COPYING.RUNTIME respectively. If not, see + <http://www.gnu.org/licenses/>. */ + +#ifndef __RISCV_TH_VECTOR_H +#define __RISCV_TH_VECTOR_H + +#include <stdint.h> +#include <stddef.h> + +#ifndef __riscv_xtheadvector +#error "XTheadVector intrinsics require the xtheadvector extension." +#else +#ifdef __cplusplus +extern "C" { +#endif + +/* NOTE: This implementation of riscv_th_vector.h is intentionally short. It does + not define the RVV types and intrinsic functions directly in C and C++ + code, but instead uses the following pragma to tell GCC to insert the + necessary type and function definitions itself. The net effect is the + same, and the file is a complete implementation of riscv_th_vector.h. */ +#pragma riscv intrinsic "vector" + +#ifdef __cplusplus +} +#endif // __cplusplus +#endif // __riscv_xtheadvector +#endif // __RISCV_TH_ECTOR_H diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md new file mode 100644 index 00000000000..1230bfb5994 --- /dev/null +++ b/gcc/config/riscv/thead-vector.md @@ -0,0 +1,69 @@ +(define_c_enum "unspec" [ + UNSPEC_TH_VWLDST +]) + +(define_mode_iterator V_VLS_VT [V VLS VT]) +(define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) + +(define_split + [(set (match_operand:V_VB_VLS_VT 0 "reg_or_mem_operand") + (match_operand:V_VB_VLS_VT 1 "reg_or_mem_operand"))] + "TARGET_XTHEADVECTOR" + [(const_int 0)] + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + }) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:V_VLS_VT 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:V_VLS_VT + [(match_operand:V_VLS_VT 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2")]) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:VB 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:VB + [(match_operand:VB 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2") + (set (attr "sew") (const_int 8)) + (set (attr "vlmul") (symbol_ref "riscv_vector::LMUL_1"))]) diff --git a/gcc/config/riscv/vector-iterators.md b/gcc/config/riscv/vector-iterators.md index 5f5f7b5b986..c0fc7a2441d 100644 --- a/gcc/config/riscv/vector-iterators.md +++ b/gcc/config/riscv/vector-iterators.md @@ -109,11 +109,11 @@ ]) (define_mode_iterator VI [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -128,11 +128,11 @@ ;; allow the instruction and mode to be matched during combine et al. (define_mode_iterator VF [ (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -140,11 +140,11 @@ (define_mode_iterator VF_ZVFHMIN [ (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -271,16 +271,16 @@ ]) (define_mode_iterator VEEWEXT2 [ - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -290,10 +290,10 @@ ]) (define_mode_iterator VEEWEXT4 [ - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -311,59 +311,59 @@ ]) (define_mode_iterator VEEWTRUNC2 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") (RVVM4SI "TARGET_64BIT") (RVVM2SI "TARGET_64BIT") (RVVM1SI "TARGET_64BIT") - (RVVMF2SI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEEWTRUNC4 [ - RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM2HI "TARGET_64BIT") (RVVM1HI "TARGET_64BIT") - (RVVMF2HI "TARGET_64BIT") - (RVVMF4HI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2HI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") (RVVM1HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEEWTRUNC8 [ (RVVM1QI "TARGET_64BIT") - (RVVMF2QI "TARGET_64BIT") - (RVVMF4QI "TARGET_64BIT") - (RVVMF8QI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2QI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF4QI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEI16 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -452,11 +452,11 @@ ]) (define_mode_iterator VFULLI [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_FULL_V") (RVVM4DI "TARGET_FULL_V") (RVVM2DI "TARGET_FULL_V") (RVVM1DI "TARGET_FULL_V") @@ -509,17 +509,17 @@ ]) (define_mode_iterator VI_QH [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") ]) (define_mode_iterator VI_QHS [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -560,11 +560,11 @@ ]) (define_mode_iterator VI_QHS_NO_M8 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -603,11 +603,11 @@ (define_mode_iterator VF_HS [ (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (V1HF "riscv_vector::vls_mode_valid_p (V1HFmode) && TARGET_ZVFH") (V2HF "riscv_vector::vls_mode_valid_p (V2HFmode) && TARGET_ZVFH") @@ -638,12 +638,12 @@ (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") - (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (V1HF "riscv_vector::vls_mode_valid_p (V1HFmode) && TARGET_ZVFH") (V2HF "riscv_vector::vls_mode_valid_p (V2HFmode) && TARGET_ZVFH") @@ -674,11 +674,11 @@ ]) (define_mode_iterator V_VLSI_QHS [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -756,27 +756,27 @@ ;; E.g. when index mode = RVVM8QImde and Pmode = SImode, if it is not zero_extend or ;; scalar != 1, such gather/scatter is not allowed since we don't have RVVM32SImode. (define_mode_iterator RATIO64 [ - (RVVMF8QI "TARGET_MIN_VLEN > 32") - (RVVMF4HI "TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM1DI "TARGET_VECTOR_ELEN_64") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") ]) (define_mode_iterator RATIO32 [ - RVVMF4QI - RVVMF2HI + (RVVMF4QI "!TARGET_XTHEADVECTOR") + (RVVMF2HI "!TARGET_XTHEADVECTOR") RVVM1SI (RVVM2DI "TARGET_VECTOR_ELEN_64") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") ]) (define_mode_iterator RATIO16 [ - RVVMF2QI + (RVVMF2QI "!TARGET_XTHEADVECTOR") RVVM1HI RVVM2SI (RVVM4DI "TARGET_VECTOR_ELEN_64") @@ -814,21 +814,21 @@ ]) (define_mode_iterator RATIO64I [ - (RVVMF8QI "TARGET_MIN_VLEN > 32") - (RVVMF4HI "TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM1DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") ]) (define_mode_iterator RATIO32I [ - RVVMF4QI - RVVMF2HI + (RVVMF4QI "!TARGET_XTHEADVECTOR") + (RVVMF2HI "!TARGET_XTHEADVECTOR") RVVM1SI (RVVM2DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") ]) (define_mode_iterator RATIO16I [ - RVVMF2QI + (RVVMF2QI "!TARGET_XTHEADVECTOR") RVVM1HI RVVM2SI (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") @@ -873,21 +873,21 @@ ]) (define_mode_iterator V_FRACT [ - RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") ]) (define_mode_iterator VWEXTI [ - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -933,7 +933,7 @@ (RVVM4SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -966,7 +966,7 @@ (RVVM4SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -996,7 +996,7 @@ (define_mode_iterator VWCONVERTI [ (RVVM8SI "TARGET_ZVFH") (RVVM4SI "TARGET_ZVFH") (RVVM2SI "TARGET_ZVFH") (RVVM1SI "TARGET_ZVFH") - (RVVMF2SI "TARGET_ZVFH") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_ZVFH") (RVVM8DI "TARGET_VECTOR_ELEN_64 && TARGET_VECTOR_ELEN_FP_32") (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_VECTOR_ELEN_FP_32") @@ -1045,7 +1045,7 @@ ]) (define_mode_iterator VQEXTI [ - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -1456,11 +1456,11 @@ ;; VINDEXED [VI8 VI16 VI32 (VI64 "TARGET_64BIT")]. (define_mode_iterator VINDEXED [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") @@ -1468,12 +1468,12 @@ (RVVM1DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64 && TARGET_64BIT") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64 && TARGET_64BIT") @@ -3173,11 +3173,11 @@ (define_mode_iterator V_VLS_F_CONVERT_SI [ (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") - (RVVMF2HF "TARGET_ZVFH") (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") @@ -3290,12 +3290,12 @@ ]) (define_mode_iterator V_VLS_F_CONVERT_DI [ - (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index b5a9055cdc4..cb30c9ae97c 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -83,7 +83,7 @@ ;; check. However, we need default value of SEW for vsetvl instruction since there ;; is no field for ratio in the vsetvl instruction encoding. (define_attr "sew" "" - (cond [(eq_attr "mode" "RVVMF64BI,RVVMF32BI,RVVMF16BI,RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ + (cond [(eq_attr "mode" "RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ RVVM8QI,RVVM4QI,RVVM2QI,RVVM1QI,RVVMF2QI,RVVMF4QI,RVVMF8QI,\ RVVM1x8QI,RVVMF2x8QI,RVVMF4x8QI,RVVMF8x8QI,\ RVVM1x7QI,RVVMF2x7QI,RVVMF4x7QI,RVVMF8x7QI,\ @@ -95,6 +95,18 @@ V1QI,V2QI,V4QI,V8QI,V16QI,V32QI,V64QI,V128QI,V256QI,V512QI,V1024QI,V2048QI,V4096QI,\ V1BI,V2BI,V4BI,V8BI,V16BI,V32BI,V64BI,V128BI,V256BI,V512BI,V1024BI,V2048BI,V4096BI") (const_int 8) + (eq_attr "mode" "RVVMF16BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 16) + (const_int 8)) + (eq_attr "mode" "RVVMF32BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 32) + (const_int 8)) + (eq_attr "mode" "RVVMF64BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 64) + (const_int 8)) (eq_attr "mode" "RVVM8HI,RVVM4HI,RVVM2HI,RVVM1HI,RVVMF2HI,RVVMF4HI,\ RVVM1x8HI,RVVMF2x8HI,RVVMF4x8HI,\ RVVM1x7HI,RVVMF2x7HI,RVVMF4x7HI,\ @@ -155,9 +167,9 @@ (eq_attr "mode" "RVVM4QI,RVVMF2BI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2QI,RVVMF4BI") (symbol_ref "riscv_vector::LMUL_2") (eq_attr "mode" "RVVM1QI,RVVMF8BI") (symbol_ref "riscv_vector::LMUL_1") - (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "riscv_vector::LMUL_F2") - (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "riscv_vector::LMUL_F4") - (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "riscv_vector::LMUL_F8") + (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F2") + (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F4") + (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F8") (eq_attr "mode" "RVVM8HI") (symbol_ref "riscv_vector::LMUL_8") (eq_attr "mode" "RVVM4HI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2HI") (symbol_ref "riscv_vector::LMUL_2") @@ -428,6 +440,10 @@ vislide1up,vislide1down,vfslide1up,vfslide1down,\ vgather,vcompress,vlsegdux,vlsegdox,vssegtux,vssegtox") (const_int INVALID_ATTRIBUTE) + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE) (eq_attr "mode" "RVVM8QI,RVVM1BI") (const_int 1) (eq_attr "mode" "RVVM4QI,RVVMF2BI") (const_int 2) (eq_attr "mode" "RVVM2QI,RVVMF4BI") (const_int 4) @@ -888,6 +904,8 @@ (symbol_ref "riscv_vector::FRM_DYN")] (symbol_ref "riscv_vector::FRM_NONE"))) +(include "thead-vector.md") + ;; ----------------------------------------------------------------- ;; ---- Miscellaneous Operations ;; ----------------------------------------------------------------- @@ -1061,6 +1079,12 @@ - We can not leave it to TARGET_SECONDARY_RELOAD since it happens before spilling. The clobber scratch is used by spilling fractional registers in IRA/LRA so it's too early. */ + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; @@ -1097,7 +1121,7 @@ (define_insn "*mov<mode>_whole" [(set (match_operand:V_WHOLE 0 "reg_or_mem_operand" "=vr, m,vr") (match_operand:V_WHOLE 1 "reg_or_mem_operand" " m,vr,vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "@ vl%m1re<sew>.v\t%0,%1 vs%m1r.v\t%1,%0 @@ -1118,6 +1142,13 @@ (match_operand:VB 1 "general_operand"))] "TARGET_VECTOR" { + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } + if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; }) @@ -1125,7 +1156,7 @@ (define_insn "*mov<mode>" [(set (match_operand:VB 0 "register_operand" "=vr") (match_operand:VB 1 "register_operand" " vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "vmv1r.v\t%0,%1" [(set_attr "type" "vmov") (set_attr "mode" "<MODE>")]) @@ -1511,7 +1542,7 @@ (match_dup 4) (match_dup 5)] UNSPEC_VSETVL))] "TARGET_VECTOR" - "vset%i1vli\t%0,%1,e%2,%m3,t%p4,m%p5" + { return TARGET_XTHEADVECTOR ? "vsetvli\t%0,%1,e%2,%m3" : "vset%i1vli\t%0,%1,e%2,%m3,t%p4,m%p5"; } [(set_attr "type" "vsetvl") (set_attr "mode" "<MODE>") (set (attr "sew") (symbol_ref "INTVAL (operands[2])")) @@ -1529,7 +1560,7 @@ (match_operand 2 "const_int_operand" "i") (match_operand 3 "const_int_operand" "i")] UNSPEC_VSETVL))] "TARGET_VECTOR" - "vsetvli\tzero,zero,e%0,%m1,t%p2,m%p3" + { return TARGET_XTHEADVECTOR ? "vsetvli\tzero,zero,e%0,%m1" : "vsetvli\tzero,zero,e%0,%m1,t%p2,m%p3"; } [(set_attr "type" "vsetvl") (set_attr "mode" "SI") (set (attr "sew") (symbol_ref "INTVAL (operands[0])")) @@ -1551,7 +1582,7 @@ (match_operand 3 "const_int_operand" "i") (match_operand 4 "const_int_operand" "i")] UNSPEC_VSETVL))] "TARGET_VECTOR" - "vset%i0vli\tzero,%0,e%1,%m2,t%p3,m%p4" + { return TARGET_XTHEADVECTOR ? "vsetvli\tzero,%0,e%1,%m2" : "vset%i0vli\tzero,%0,e%1,%m2,t%p3,m%p4"; } [(set_attr "type" "vsetvl") (set_attr "mode" "<MODE>") (set (attr "sew") (symbol_ref "INTVAL (operands[1])")) @@ -3680,7 +3711,7 @@ (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3701,7 +3732,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3722,7 +3753,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c index 2e0e12aa045..2eef9e1e1a8 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c @@ -1,4 +1,4 @@ -/* { dg-do compile } */ +/* { dg-do compile { target { ! riscv_xtheadvector } } } */ /* { dg-skip-if "test rvv intrinsic" { *-*-* } { "*" } { "-march=rv*v*" } } */ void foo0 () {__rvv_bool64_t t;} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c index 3d81b179235..ef329e30785 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c @@ -1,4 +1,4 @@ /* { dg-do compile } */ /* { dg-options "-O3 -march=rv32gc -mabi=ilp32d" } */ -#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' extension enabled} } */ +#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' or 'XTHEADVECTOR' extension enabled} } */ diff --git a/gcc/testsuite/lib/target-supports.exp b/gcc/testsuite/lib/target-supports.exp index 7f13ff0ca56..70df6b1401c 100644 --- a/gcc/testsuite/lib/target-supports.exp +++ b/gcc/testsuite/lib/target-supports.exp @@ -1952,6 +1952,18 @@ proc check_effective_target_riscv_zbb { } { }] } +# Return 1 if the target arch supports the XTheadVector extension, 0 otherwise. +# Cache the result. + +proc check_effective_target_riscv_xtheadvector { } { + return [check_no_compiler_messages riscv_ext_xtheadvector assembly { + #ifndef __riscv_xtheadvector + #error "Not __riscv_xtheadvector" + #endif + }] +} + + # Return 1 if we can execute code when using dg-add-options riscv_v proc check_effective_target_riscv_v_ok { } { -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: [PATCH v4] RISC-V: Handle differences between XTheadvector and Vector 2024-01-02 12:39 ` [PATCH " Jun Sha (Joshua) @ 2024-01-03 3:11 ` Kito Cheng 2024-01-03 6:15 ` Jun Sha (Joshua) 1 sibling, 0 replies; 133+ messages in thread From: Kito Cheng @ 2024-01-03 3:11 UTC (permalink / raw) To: Jun Sha (Joshua) Cc: gcc-patches, jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, Jin Ma, Xianmiao Qu > diff --git a/gcc/config/riscv/riscv_th_vector.h b/gcc/config/riscv/riscv_th_vector.h > new file mode 100644 > index 00000000000..6f47e0c90a4 > --- /dev/null > +++ b/gcc/config/riscv/riscv_th_vector.h > @@ -0,0 +1,49 @@ > +/* RISC-V 'XTheadVector' Extension intrinsics include file. > + Copyright (C) 2022-2023 Free Software Foundation, Inc. > + > + This file is part of GCC. > + > + GCC is free software; you can redistribute it and/or modify it > + under the terms of the GNU General Public License as published > + by the Free Software Foundation; either version 3, or (at your > + option) any later version. > + > + GCC is distributed in the hope that it will be useful, but WITHOUT > + ANY WARRANTY; without even the implied warranty of MERCHANTABILITY > + or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public > + License for more details. > + > + Under Section 7 of GPL version 3, you are granted additional > + permissions described in the GCC Runtime Library Exception, version > + 3.1, as published by the Free Software Foundation. > + > + You should have received a copy of the GNU General Public License and > + a copy of the GCC Runtime Library Exception along with this program; > + see the files COPYING3 and COPYING.RUNTIME respectively. If not, see > + <http://www.gnu.org/licenses/>. */ > + > +#ifndef __RISCV_TH_VECTOR_H > +#define __RISCV_TH_VECTOR_H > + > +#include <stdint.h> > +#include <stddef.h> > + > +#ifndef __riscv_xtheadvector > +#error "XTheadVector intrinsics require the xtheadvector extension." > +#else > +#ifdef __cplusplus > +extern "C" { > +#endif > + > +/* NOTE: This implementation of riscv_th_vector.h is intentionally short. It does > + not define the RVV types and intrinsic functions directly in C and C++ > + code, but instead uses the following pragma to tell GCC to insert the > + necessary type and function definitions itself. The net effect is the > + same, and the file is a complete implementation of riscv_th_vector.h. */ > +#pragma riscv intrinsic "vector" #pragma riscv intrinsic "theadvector" Don't reuse `#pragma riscv intrinsic "vector"` to prevent including riscv_vector.h work with __riscv_xtheadvector. I know we already guarded with ifndef __riscv_xtheadvector and ifndef __riscv_vector for now, but we eventually will remove that due to multi-version function support. e.g. a.c compile with -march=rv64gc a.c: #include <riscv_vector.h> void foo(){ ... } void foo_vector () __attribute__(("arch=+v")); void foo_vector () { // Use vector intrinsic to implement something } ^ permalink raw reply [flat|nested] 133+ messages in thread
* [PATCH v4] RISC-V: Handle differences between XTheadvector and Vector 2024-01-02 12:39 ` [PATCH " Jun Sha (Joshua) 2024-01-03 3:11 ` Kito Cheng @ 2024-01-03 6:15 ` Jun Sha (Joshua) 2024-01-04 2:29 ` Jun Sha (Joshua) 2024-01-08 23:08 ` [PATCH v4] " 钟居哲 1 sibling, 2 replies; 133+ messages in thread From: Jun Sha (Joshua) @ 2024-01-03 6:15 UTC (permalink / raw) To: gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, Jun Sha (Joshua), Jin Ma, Xianmiao Qu This patch is to handle the differences in instruction generation between Vector and XTheadVector. In this version, we only support partial xtheadvector instructions that leverage directly from current RVV1.0 with simple adding "th." prefix. For different name xtheadvector instructions but share same patterns as RVV1.0 instructions, we will use ASM targethook to rewrite the whole string of the instructions in the following patches. For some vector patterns that cannot be avoided, we use "!TARGET_XTHEADVECTOR" to disable them in vector.md in order not to generate instructions that xtheadvector does not support, like vmv1r and vsext.vf2. gcc/ChangeLog: * config.gcc: Add files for XTheadVector intrinsics. * config/riscv/autovec.md: Guard XTheadVector. * config/riscv/riscv-c.cc: Add pragma for XTheadVector. * config/riscv/riscv-string.cc (expand_block_move): Guard XTheadVector. (get_prefer_tail_policy): Give specific value for tail. (get_prefer_mask_policy): Give specific value for mask. (vls_mode_valid_p): Avoid autovec. * config/riscv/riscv-vector-builtins-shapes.cc (check_type): (build_one): New function. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): (DEF_THEAD_RVV_FUNCTION): Add new marcos. (check_required_extensions): (handle_pragma_vector): * config/riscv/riscv-vector-builtins.h (RVV_REQUIRE_VECTOR): (RVV_REQUIRE_XTHEADVECTOR): Add RVV_REQUIRE_VECTOR and RVV_REQUIRE_XTHEADVECTOR. (struct function_group_info): * config/riscv/riscv-vector-switch.def (ENTRY): Disable fractional mode for the XTheadVector extension. (TUPLE_ENTRY): Likewise. * config/riscv/riscv-vsetvl.cc: Add functions for xtheadvector. * config/riscv/riscv.cc (riscv_v_ext_vls_mode_p): Guard XTheadVector. (riscv_v_adjust_bytesize): Likewise. (riscv_preferred_simd_mode): Likewsie. (riscv_autovectorize_vector_modes): Likewise. (riscv_vector_mode_supported_any_target_p): Likewise. (TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P): Likewise. * config/riscv/vector-iterators.md: Remove fractional LMUL. * config/riscv/vector.md: Include thead-vector.md. * config/riscv/riscv_th_vector.h: New file. * config/riscv/thead-vector.md: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/base/pragma-1.c: Add XTheadVector. * gcc.target/riscv/rvv/base/abi-1.c: Exclude XTheadVector. * lib/target-supports.exp: Add target for XTheadVector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config.gcc | 2 +- gcc/config/riscv/autovec.md | 2 +- gcc/config/riscv/predicates.md | 4 +- gcc/config/riscv/riscv-c.cc | 3 +- gcc/config/riscv/riscv-string.cc | 3 +- gcc/config/riscv/riscv-v.cc | 6 +- .../riscv/riscv-vector-builtins-bases.cc | 48 +++-- .../riscv/riscv-vector-builtins-shapes.cc | 23 +++ gcc/config/riscv/riscv-vector-switch.def | 150 +++++++------- gcc/config/riscv/riscv.cc | 20 +- gcc/config/riscv/riscv_th_vector.h | 49 +++++ gcc/config/riscv/thead-vector.md | 69 +++++++ gcc/config/riscv/vector-iterators.md | 186 +++++++++--------- gcc/config/riscv/vector.md | 55 ++++-- .../gcc.target/riscv/rvv/base/abi-1.c | 2 +- .../gcc.target/riscv/rvv/base/pragma-1.c | 2 +- gcc/testsuite/lib/target-supports.exp | 12 ++ 17 files changed, 427 insertions(+), 209 deletions(-) create mode 100644 gcc/config/riscv/riscv_th_vector.h create mode 100644 gcc/config/riscv/thead-vector.md diff --git a/gcc/config.gcc b/gcc/config.gcc index f0676c830e8..1445d98c147 100644 --- a/gcc/config.gcc +++ b/gcc/config.gcc @@ -549,7 +549,7 @@ riscv*) extra_objs="${extra_objs} riscv-vector-builtins.o riscv-vector-builtins-shapes.o riscv-vector-builtins-bases.o" extra_objs="${extra_objs} thead.o riscv-target-attr.o" d_target_objs="riscv-d.o" - extra_headers="riscv_vector.h" + extra_headers="riscv_vector.h riscv_th_vector.h" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.cc" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.h" ;; diff --git a/gcc/config/riscv/autovec.md b/gcc/config/riscv/autovec.md index 8b8a92f10a1..1fac56c7095 100644 --- a/gcc/config/riscv/autovec.md +++ b/gcc/config/riscv/autovec.md @@ -2579,7 +2579,7 @@ [(match_operand 0 "register_operand") (match_operand 1 "memory_operand") (match_operand:ANYI 2 "const_int_operand")] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" { riscv_vector::expand_rawmemchr(<MODE>mode, operands[0], operands[1], operands[2]); diff --git a/gcc/config/riscv/predicates.md b/gcc/config/riscv/predicates.md index b87a6900841..6e36c5afae1 100644 --- a/gcc/config/riscv/predicates.md +++ b/gcc/config/riscv/predicates.md @@ -428,7 +428,9 @@ ;; Predicates for the V extension. (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (ior (match_test "TARGET_XTHEADVECTOR && rtx_equal_p (op, const0_rtx)") + (match_test "!TARGET_XTHEADVECTOR")) + (match_operand 0 "const_csr_operand")))) (define_special_predicate "autovec_length_operand" (ior (match_operand 0 "pmode_register_operand") diff --git a/gcc/config/riscv/riscv-c.cc b/gcc/config/riscv/riscv-c.cc index d290c58613a..d6fb83ac00c 100644 --- a/gcc/config/riscv/riscv-c.cc +++ b/gcc/config/riscv/riscv-c.cc @@ -195,7 +195,8 @@ riscv_pragma_intrinsic (cpp_reader *) const char *name = TREE_STRING_POINTER (x); - if (strcmp (name, "vector") == 0) + if (strcmp (name, "vector") == 0 + || strcmp (name, "xtheadvector") == 0) { if (!TARGET_VECTOR) { diff --git a/gcc/config/riscv/riscv-string.cc b/gcc/config/riscv/riscv-string.cc index 11c1f74d0b3..f431b562222 100644 --- a/gcc/config/riscv/riscv-string.cc +++ b/gcc/config/riscv/riscv-string.cc @@ -773,7 +773,8 @@ riscv_expand_block_move_scalar (rtx dest, rtx src, rtx length) bool riscv_expand_block_move (rtx dest, rtx src, rtx length) { - if (TARGET_VECTOR && stringop_strategy & STRATEGY_VECTOR) + if ((TARGET_VECTOR && !TARGET_XTHEADVECTOR) + && stringop_strategy & STRATEGY_VECTOR) { bool ok = riscv_vector::expand_block_move (dest, src, length); if (ok) diff --git a/gcc/config/riscv/riscv-v.cc b/gcc/config/riscv/riscv-v.cc index 038ab084a37..eb061b593a5 100644 --- a/gcc/config/riscv/riscv-v.cc +++ b/gcc/config/riscv/riscv-v.cc @@ -1772,7 +1772,7 @@ get_prefer_tail_policy () compiler pick up either agnostic or undisturbed. Maybe we will have a compile option like -mprefer=agnostic to set this value???. */ - return TAIL_ANY; + return TARGET_XTHEADVECTOR ? TAIL_AGNOSTIC : TAIL_ANY; } /* Get prefer mask policy. */ @@ -1783,7 +1783,7 @@ get_prefer_mask_policy () compiler pick up either agnostic or undisturbed. Maybe we will have a compile option like -mprefer=agnostic to set this value???. */ - return MASK_ANY; + return TARGET_XTHEADVECTOR ? MASK_UNDISTURBED : MASK_ANY; } /* Get avl_type rtx. */ @@ -4383,7 +4383,7 @@ cmp_lmul_gt_one (machine_mode mode) bool vls_mode_valid_p (machine_mode vls_mode) { - if (!TARGET_VECTOR) + if (!TARGET_VECTOR || TARGET_XTHEADVECTOR) return false; if (riscv_autovec_preference == RVV_SCALABLE) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index c51affde353..22d877af9be 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -115,23 +115,37 @@ public: tree type = builtin_types[e.type.index].vector; machine_mode mode = TYPE_MODE (type); - /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. - - - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 - - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 - - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 - - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 - - e8,m2/e16,m4/e32,m8 --> e8m2 - - e8,m4/e16,m8 --> e8m4 - - e8,m8 --> e8m8 - */ - /* SEW. */ - e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); - - /* LMUL. */ - machine_mode e8_mode - = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); - e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + + if (TARGET_XTHEADVECTOR) + { + machine_mode inner_mode = GET_MODE_INNER (mode); + /* SEW. */ + e.add_input_operand (Pmode, + gen_int_mode (GET_MODE_BITSIZE (inner_mode), Pmode)); + /* LMUL. */ + e.add_input_operand (Pmode, + gen_int_mode (get_vlmul (mode), Pmode)); + } + else + { + /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. + + - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 + - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 + - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 + - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 + - e8,m2/e16,m4/e32,m8 --> e8m2 + - e8,m4/e16,m8 --> e8m4 + - e8,m8 --> e8m8 + */ + /* SEW. */ + e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); + + /* LMUL. */ + machine_mode e8_mode + = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); + e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + } /* TAIL_ANY. */ e.add_input_operand (Pmode, diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 4a754e0228f..6b49404a1fa 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -33,6 +33,25 @@ namespace riscv_vector { +/* Check whether the RETURN_TYPE and ARGUMENT_TYPES are + valid for the function. */ + +static bool +check_type (tree return_type, vec<tree> &argument_types) +{ + tree arg; + unsigned i; + + if (!return_type) + return false; + + FOR_EACH_VEC_ELT (argument_types, i, arg) + if (!arg) + return false; + + return true; +} + /* Add one function instance for GROUP, using operand suffix at index OI, mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ static void @@ -49,6 +68,10 @@ build_one (function_builder &b, const function_group_info &group, group.ops_infos.types[vec_type_idx].index); b.allocate_argument_types (function_instance, argument_types); b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + b.add_overloaded_function (function_instance, *group.shape); b.add_unique_function (function_instance, (*group.shape), return_type, argument_types); diff --git a/gcc/config/riscv/riscv-vector-switch.def b/gcc/config/riscv/riscv-vector-switch.def index 5c9f9bcbc3e..f7a66b34bae 100644 --- a/gcc/config/riscv/riscv-vector-switch.def +++ b/gcc/config/riscv/riscv-vector-switch.def @@ -68,9 +68,9 @@ Encode the ratio of SEW/LMUL into the mask types. #endif /* Disable modes if TARGET_MIN_VLEN == 32. */ -ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) -ENTRY (RVVMF32BI, true, LMUL_F4, 32) -ENTRY (RVVMF16BI, true, LMUL_F2, 16) +ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F8, 64) +ENTRY (RVVMF32BI, true, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F4, 32) +ENTRY (RVVMF16BI, true, TARGET_XTHEADVECTOR ? LMUL_1 : LMUL_F2 , 16) ENTRY (RVVMF8BI, true, LMUL_1, 8) ENTRY (RVVMF4BI, true, LMUL_2, 4) ENTRY (RVVMF2BI, true, LMUL_4, 2) @@ -81,39 +81,39 @@ ENTRY (RVVM8QI, true, LMUL_8, 1) ENTRY (RVVM4QI, true, LMUL_4, 2) ENTRY (RVVM2QI, true, LMUL_2, 4) ENTRY (RVVM1QI, true, LMUL_1, 8) -ENTRY (RVVMF2QI, true, LMUL_F2, 16) -ENTRY (RVVMF4QI, true, LMUL_F4, 32) -ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) +ENTRY (RVVMF2QI, !TARGET_XTHEADVECTOR, LMUL_F2, 16) +ENTRY (RVVMF4QI, !TARGET_XTHEADVECTOR, LMUL_F4, 32) +ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F8, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8HI, true, LMUL_8, 2) ENTRY (RVVM4HI, true, LMUL_4, 4) ENTRY (RVVM2HI, true, LMUL_2, 8) ENTRY (RVVM1HI, true, LMUL_1, 16) -ENTRY (RVVMF2HI, true, LMUL_F2, 32) -ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HI, !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_16. */ ENTRY (RVVM8HF, TARGET_VECTOR_ELEN_FP_16, LMUL_8, 2) ENTRY (RVVM4HF, TARGET_VECTOR_ELEN_FP_16, LMUL_4, 4) ENTRY (RVVM2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_2, 8) ENTRY (RVVM1HF, TARGET_VECTOR_ELEN_FP_16, LMUL_1, 16) -ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_F2, 32) -ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8SI, true, LMUL_8, 4) ENTRY (RVVM4SI, true, LMUL_4, 8) ENTRY (RVVM2SI, true, LMUL_2, 16) ENTRY (RVVM1SI, true, LMUL_1, 32) -ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_32. */ ENTRY (RVVM8SF, TARGET_VECTOR_ELEN_FP_32, LMUL_8, 4) ENTRY (RVVM4SF, TARGET_VECTOR_ELEN_FP_32, LMUL_4, 8) ENTRY (RVVM2SF, TARGET_VECTOR_ELEN_FP_32, LMUL_2, 16) ENTRY (RVVM1SF, TARGET_VECTOR_ELEN_FP_32, LMUL_1, 32) -ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if !TARGET_VECTOR_ELEN_64. */ ENTRY (RVVM8DI, TARGET_VECTOR_ELEN_64, LMUL_8, 8) @@ -140,127 +140,127 @@ ENTRY (RVVM1DF, TARGET_VECTOR_ELEN_FP_64, LMUL_1, 64) #endif TUPLE_ENTRY (RVVM1x8QI, true, RVVM1QI, 8, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x8QI, true, RVVMF2QI, 8, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x8QI, true, RVVMF4QI, 8, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 8, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x8QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 8, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x8QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 8, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 8, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x7QI, true, RVVM1QI, 7, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x7QI, true, RVVMF2QI, 7, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x7QI, true, RVVMF4QI, 7, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 7, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x7QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 7, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x7QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 7, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 7, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x6QI, true, RVVM1QI, 6, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x6QI, true, RVVMF2QI, 6, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x6QI, true, RVVMF4QI, 6, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 6, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x6QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 6, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x6QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 6, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 6, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x5QI, true, RVVM1QI, 5, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x5QI, true, RVVMF2QI, 5, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x5QI, true, RVVMF4QI, 5, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 5, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x5QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 5, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x5QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 5, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 5, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x4QI, true, RVVM2QI, 4, LMUL_2, 4) TUPLE_ENTRY (RVVM1x4QI, true, RVVM1QI, 4, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x4QI, true, RVVMF2QI, 4, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x4QI, true, RVVMF4QI, 4, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 4, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x4QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 4, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x4QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 4, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 4, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x3QI, true, RVVM2QI, 3, LMUL_2, 4) TUPLE_ENTRY (RVVM1x3QI, true, RVVM1QI, 3, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x3QI, true, RVVMF2QI, 3, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x3QI, true, RVVMF4QI, 3, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 3, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x3QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 3, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x3QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 3, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 3, LMUL_F8, 64) TUPLE_ENTRY (RVVM4x2QI, true, RVVM4QI, 2, LMUL_4, 2) TUPLE_ENTRY (RVVM2x2QI, true, RVVM2QI, 2, LMUL_2, 4) TUPLE_ENTRY (RVVM1x2QI, true, RVVM1QI, 2, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x2QI, true, RVVMF2QI, 2, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x2QI, true, RVVMF4QI, 2, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 2, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x2QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 2, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x2QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 2, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 2, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x8HI, true, RVVM1HI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HI, true, RVVMF2HI, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HI, true, RVVM1HI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HI, true, RVVMF2HI, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HI, true, RVVM1HI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HI, true, RVVMF2HI, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HI, true, RVVM1HI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HI, true, RVVMF2HI, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HI, true, RVVM2HI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HI, true, RVVM1HI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HI, true, RVVMF2HI, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HI, true, RVVM2HI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HI, true, RVVM1HI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HI, true, RVVMF2HI, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HI, true, RVVM4HI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HI, true, RVVM2HI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HI, true, RVVM1HI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HI, true, RVVMF2HI, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM4HF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8SI, true, RVVM1SI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SI, true, RVVM1SI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SI, true, RVVM1SI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SI, true, RVVM1SI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SI, true, RVVM2SI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SI, true, RVVM1SI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SI, true, RVVM2SI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SI, true, RVVM1SI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SI, true, RVVM4SI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SI, true, RVVM2SI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SI, true, RVVM1SI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM4SF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 8, LMUL_1, 16) TUPLE_ENTRY (RVVM1x7DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 7, LMUL_1, 16) diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index 51878797287..44a36db7803 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -1406,6 +1406,9 @@ riscv_v_adjust_bytesize (machine_mode mode, int scale) { if (riscv_v_ext_vector_mode_p (mode)) { + if (TARGET_XTHEADVECTOR) + return BYTES_PER_RISCV_VECTOR; + poly_int64 nunits = GET_MODE_NUNITS (mode); poly_int64 mode_size = GET_MODE_SIZE (mode); @@ -9962,7 +9965,7 @@ riscv_use_divmod_expander (void) static machine_mode riscv_preferred_simd_mode (scalar_mode mode) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::preferred_simd_mode (mode); return word_mode; @@ -10313,7 +10316,7 @@ riscv_mode_priority (int, int n) unsigned int riscv_autovectorize_vector_modes (vector_modes *modes, bool all) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::autovectorize_vector_modes (modes, all); return default_autovectorize_vector_modes (modes, all); @@ -10496,6 +10499,16 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) return false; } +/* Implements target hook vector_mode_supported_any_target_p. */ + +static bool +riscv_vector_mode_supported_any_target_p (machine_mode mode) +{ + if (TARGET_XTHEADVECTOR) + return false; + return true; +} + /* Initialize the GCC target structure. */ #undef TARGET_ASM_ALIGNED_HI_OP #define TARGET_ASM_ALIGNED_HI_OP "\t.half\t" @@ -10839,6 +10852,9 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) #undef TARGET_PREFERRED_ELSE_VALUE #define TARGET_PREFERRED_ELSE_VALUE riscv_preferred_else_value +#undef TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P +#define TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P riscv_vector_mode_supported_any_target_p + struct gcc_target targetm = TARGET_INITIALIZER; #include "gt-riscv.h" diff --git a/gcc/config/riscv/riscv_th_vector.h b/gcc/config/riscv/riscv_th_vector.h new file mode 100644 index 00000000000..a897b9cb90a --- /dev/null +++ b/gcc/config/riscv/riscv_th_vector.h @@ -0,0 +1,49 @@ +/* RISC-V 'XTheadVector' Extension intrinsics include file. + Copyright (C) 2022-2023 Free Software Foundation, Inc. + + This file is part of GCC. + + GCC is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published + by the Free Software Foundation; either version 3, or (at your + option) any later version. + + GCC is distributed in the hope that it will be useful, but WITHOUT + ANY WARRANTY; without even the implied warranty of MERCHANTABILITY + or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public + License for more details. + + Under Section 7 of GPL version 3, you are granted additional + permissions described in the GCC Runtime Library Exception, version + 3.1, as published by the Free Software Foundation. + + You should have received a copy of the GNU General Public License and + a copy of the GCC Runtime Library Exception along with this program; + see the files COPYING3 and COPYING.RUNTIME respectively. If not, see + <http://www.gnu.org/licenses/>. */ + +#ifndef __RISCV_TH_VECTOR_H +#define __RISCV_TH_VECTOR_H + +#include <stdint.h> +#include <stddef.h> + +#ifndef __riscv_xtheadvector +#error "XTheadVector intrinsics require the xtheadvector extension." +#else +#ifdef __cplusplus +extern "C" { +#endif + +/* NOTE: This implementation of riscv_th_vector.h is intentionally short. It does + not define the RVV types and intrinsic functions directly in C and C++ + code, but instead uses the following pragma to tell GCC to insert the + necessary type and function definitions itself. The net effect is the + same, and the file is a complete implementation of riscv_th_vector.h. */ +#pragma riscv intrinsic "xtheadvector" + +#ifdef __cplusplus +} +#endif // __cplusplus +#endif // __riscv_xtheadvector +#endif // __RISCV_TH_ECTOR_H diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md new file mode 100644 index 00000000000..1230bfb5994 --- /dev/null +++ b/gcc/config/riscv/thead-vector.md @@ -0,0 +1,69 @@ +(define_c_enum "unspec" [ + UNSPEC_TH_VWLDST +]) + +(define_mode_iterator V_VLS_VT [V VLS VT]) +(define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) + +(define_split + [(set (match_operand:V_VB_VLS_VT 0 "reg_or_mem_operand") + (match_operand:V_VB_VLS_VT 1 "reg_or_mem_operand"))] + "TARGET_XTHEADVECTOR" + [(const_int 0)] + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + }) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:V_VLS_VT 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:V_VLS_VT + [(match_operand:V_VLS_VT 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2")]) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:VB 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:VB + [(match_operand:VB 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2") + (set (attr "sew") (const_int 8)) + (set (attr "vlmul") (symbol_ref "riscv_vector::LMUL_1"))]) diff --git a/gcc/config/riscv/vector-iterators.md b/gcc/config/riscv/vector-iterators.md index 5f5f7b5b986..c0fc7a2441d 100644 --- a/gcc/config/riscv/vector-iterators.md +++ b/gcc/config/riscv/vector-iterators.md @@ -109,11 +109,11 @@ ]) (define_mode_iterator VI [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -128,11 +128,11 @@ ;; allow the instruction and mode to be matched during combine et al. (define_mode_iterator VF [ (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -140,11 +140,11 @@ (define_mode_iterator VF_ZVFHMIN [ (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -271,16 +271,16 @@ ]) (define_mode_iterator VEEWEXT2 [ - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -290,10 +290,10 @@ ]) (define_mode_iterator VEEWEXT4 [ - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -311,59 +311,59 @@ ]) (define_mode_iterator VEEWTRUNC2 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") (RVVM4SI "TARGET_64BIT") (RVVM2SI "TARGET_64BIT") (RVVM1SI "TARGET_64BIT") - (RVVMF2SI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEEWTRUNC4 [ - RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM2HI "TARGET_64BIT") (RVVM1HI "TARGET_64BIT") - (RVVMF2HI "TARGET_64BIT") - (RVVMF4HI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2HI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") (RVVM1HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEEWTRUNC8 [ (RVVM1QI "TARGET_64BIT") - (RVVMF2QI "TARGET_64BIT") - (RVVMF4QI "TARGET_64BIT") - (RVVMF8QI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2QI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF4QI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEI16 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -452,11 +452,11 @@ ]) (define_mode_iterator VFULLI [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_FULL_V") (RVVM4DI "TARGET_FULL_V") (RVVM2DI "TARGET_FULL_V") (RVVM1DI "TARGET_FULL_V") @@ -509,17 +509,17 @@ ]) (define_mode_iterator VI_QH [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") ]) (define_mode_iterator VI_QHS [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -560,11 +560,11 @@ ]) (define_mode_iterator VI_QHS_NO_M8 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -603,11 +603,11 @@ (define_mode_iterator VF_HS [ (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (V1HF "riscv_vector::vls_mode_valid_p (V1HFmode) && TARGET_ZVFH") (V2HF "riscv_vector::vls_mode_valid_p (V2HFmode) && TARGET_ZVFH") @@ -638,12 +638,12 @@ (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") - (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (V1HF "riscv_vector::vls_mode_valid_p (V1HFmode) && TARGET_ZVFH") (V2HF "riscv_vector::vls_mode_valid_p (V2HFmode) && TARGET_ZVFH") @@ -674,11 +674,11 @@ ]) (define_mode_iterator V_VLSI_QHS [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -756,27 +756,27 @@ ;; E.g. when index mode = RVVM8QImde and Pmode = SImode, if it is not zero_extend or ;; scalar != 1, such gather/scatter is not allowed since we don't have RVVM32SImode. (define_mode_iterator RATIO64 [ - (RVVMF8QI "TARGET_MIN_VLEN > 32") - (RVVMF4HI "TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM1DI "TARGET_VECTOR_ELEN_64") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") ]) (define_mode_iterator RATIO32 [ - RVVMF4QI - RVVMF2HI + (RVVMF4QI "!TARGET_XTHEADVECTOR") + (RVVMF2HI "!TARGET_XTHEADVECTOR") RVVM1SI (RVVM2DI "TARGET_VECTOR_ELEN_64") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") ]) (define_mode_iterator RATIO16 [ - RVVMF2QI + (RVVMF2QI "!TARGET_XTHEADVECTOR") RVVM1HI RVVM2SI (RVVM4DI "TARGET_VECTOR_ELEN_64") @@ -814,21 +814,21 @@ ]) (define_mode_iterator RATIO64I [ - (RVVMF8QI "TARGET_MIN_VLEN > 32") - (RVVMF4HI "TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM1DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") ]) (define_mode_iterator RATIO32I [ - RVVMF4QI - RVVMF2HI + (RVVMF4QI "!TARGET_XTHEADVECTOR") + (RVVMF2HI "!TARGET_XTHEADVECTOR") RVVM1SI (RVVM2DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") ]) (define_mode_iterator RATIO16I [ - RVVMF2QI + (RVVMF2QI "!TARGET_XTHEADVECTOR") RVVM1HI RVVM2SI (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") @@ -873,21 +873,21 @@ ]) (define_mode_iterator V_FRACT [ - RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") ]) (define_mode_iterator VWEXTI [ - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -933,7 +933,7 @@ (RVVM4SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -966,7 +966,7 @@ (RVVM4SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -996,7 +996,7 @@ (define_mode_iterator VWCONVERTI [ (RVVM8SI "TARGET_ZVFH") (RVVM4SI "TARGET_ZVFH") (RVVM2SI "TARGET_ZVFH") (RVVM1SI "TARGET_ZVFH") - (RVVMF2SI "TARGET_ZVFH") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_ZVFH") (RVVM8DI "TARGET_VECTOR_ELEN_64 && TARGET_VECTOR_ELEN_FP_32") (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_VECTOR_ELEN_FP_32") @@ -1045,7 +1045,7 @@ ]) (define_mode_iterator VQEXTI [ - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -1456,11 +1456,11 @@ ;; VINDEXED [VI8 VI16 VI32 (VI64 "TARGET_64BIT")]. (define_mode_iterator VINDEXED [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") @@ -1468,12 +1468,12 @@ (RVVM1DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64 && TARGET_64BIT") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64 && TARGET_64BIT") @@ -3173,11 +3173,11 @@ (define_mode_iterator V_VLS_F_CONVERT_SI [ (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") - (RVVMF2HF "TARGET_ZVFH") (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") @@ -3290,12 +3290,12 @@ ]) (define_mode_iterator V_VLS_F_CONVERT_DI [ - (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index b5a9055cdc4..cb30c9ae97c 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -83,7 +83,7 @@ ;; check. However, we need default value of SEW for vsetvl instruction since there ;; is no field for ratio in the vsetvl instruction encoding. (define_attr "sew" "" - (cond [(eq_attr "mode" "RVVMF64BI,RVVMF32BI,RVVMF16BI,RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ + (cond [(eq_attr "mode" "RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ RVVM8QI,RVVM4QI,RVVM2QI,RVVM1QI,RVVMF2QI,RVVMF4QI,RVVMF8QI,\ RVVM1x8QI,RVVMF2x8QI,RVVMF4x8QI,RVVMF8x8QI,\ RVVM1x7QI,RVVMF2x7QI,RVVMF4x7QI,RVVMF8x7QI,\ @@ -95,6 +95,18 @@ V1QI,V2QI,V4QI,V8QI,V16QI,V32QI,V64QI,V128QI,V256QI,V512QI,V1024QI,V2048QI,V4096QI,\ V1BI,V2BI,V4BI,V8BI,V16BI,V32BI,V64BI,V128BI,V256BI,V512BI,V1024BI,V2048BI,V4096BI") (const_int 8) + (eq_attr "mode" "RVVMF16BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 16) + (const_int 8)) + (eq_attr "mode" "RVVMF32BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 32) + (const_int 8)) + (eq_attr "mode" "RVVMF64BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 64) + (const_int 8)) (eq_attr "mode" "RVVM8HI,RVVM4HI,RVVM2HI,RVVM1HI,RVVMF2HI,RVVMF4HI,\ RVVM1x8HI,RVVMF2x8HI,RVVMF4x8HI,\ RVVM1x7HI,RVVMF2x7HI,RVVMF4x7HI,\ @@ -155,9 +167,9 @@ (eq_attr "mode" "RVVM4QI,RVVMF2BI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2QI,RVVMF4BI") (symbol_ref "riscv_vector::LMUL_2") (eq_attr "mode" "RVVM1QI,RVVMF8BI") (symbol_ref "riscv_vector::LMUL_1") - (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "riscv_vector::LMUL_F2") - (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "riscv_vector::LMUL_F4") - (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "riscv_vector::LMUL_F8") + (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F2") + (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F4") + (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F8") (eq_attr "mode" "RVVM8HI") (symbol_ref "riscv_vector::LMUL_8") (eq_attr "mode" "RVVM4HI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2HI") (symbol_ref "riscv_vector::LMUL_2") @@ -428,6 +440,10 @@ vislide1up,vislide1down,vfslide1up,vfslide1down,\ vgather,vcompress,vlsegdux,vlsegdox,vssegtux,vssegtox") (const_int INVALID_ATTRIBUTE) + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE) (eq_attr "mode" "RVVM8QI,RVVM1BI") (const_int 1) (eq_attr "mode" "RVVM4QI,RVVMF2BI") (const_int 2) (eq_attr "mode" "RVVM2QI,RVVMF4BI") (const_int 4) @@ -888,6 +904,8 @@ (symbol_ref "riscv_vector::FRM_DYN")] (symbol_ref "riscv_vector::FRM_NONE"))) +(include "thead-vector.md") + ;; ----------------------------------------------------------------- ;; ---- Miscellaneous Operations ;; ----------------------------------------------------------------- @@ -1061,6 +1079,12 @@ - We can not leave it to TARGET_SECONDARY_RELOAD since it happens before spilling. The clobber scratch is used by spilling fractional registers in IRA/LRA so it's too early. */ + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; @@ -1097,7 +1121,7 @@ (define_insn "*mov<mode>_whole" [(set (match_operand:V_WHOLE 0 "reg_or_mem_operand" "=vr, m,vr") (match_operand:V_WHOLE 1 "reg_or_mem_operand" " m,vr,vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "@ vl%m1re<sew>.v\t%0,%1 vs%m1r.v\t%1,%0 @@ -1118,6 +1142,13 @@ (match_operand:VB 1 "general_operand"))] "TARGET_VECTOR" { + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } + if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; }) @@ -1125,7 +1156,7 @@ (define_insn "*mov<mode>" [(set (match_operand:VB 0 "register_operand" "=vr") (match_operand:VB 1 "register_operand" " vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "vmv1r.v\t%0,%1" [(set_attr "type" "vmov") (set_attr "mode" "<MODE>")]) @@ -1511,7 +1542,7 @@ (match_dup 4) (match_dup 5)] UNSPEC_VSETVL))] "TARGET_VECTOR" - "vset%i1vli\t%0,%1,e%2,%m3,t%p4,m%p5" + { return TARGET_XTHEADVECTOR ? "vsetvli\t%0,%1,e%2,%m3" : "vset%i1vli\t%0,%1,e%2,%m3,t%p4,m%p5"; } [(set_attr "type" "vsetvl") (set_attr "mode" "<MODE>") (set (attr "sew") (symbol_ref "INTVAL (operands[2])")) @@ -1529,7 +1560,7 @@ (match_operand 2 "const_int_operand" "i") (match_operand 3 "const_int_operand" "i")] UNSPEC_VSETVL))] "TARGET_VECTOR" - "vsetvli\tzero,zero,e%0,%m1,t%p2,m%p3" + { return TARGET_XTHEADVECTOR ? "vsetvli\tzero,zero,e%0,%m1" : "vsetvli\tzero,zero,e%0,%m1,t%p2,m%p3"; } [(set_attr "type" "vsetvl") (set_attr "mode" "SI") (set (attr "sew") (symbol_ref "INTVAL (operands[0])")) @@ -1551,7 +1582,7 @@ (match_operand 3 "const_int_operand" "i") (match_operand 4 "const_int_operand" "i")] UNSPEC_VSETVL))] "TARGET_VECTOR" - "vset%i0vli\tzero,%0,e%1,%m2,t%p3,m%p4" + { return TARGET_XTHEADVECTOR ? "vsetvli\tzero,%0,e%1,%m2" : "vset%i0vli\tzero,%0,e%1,%m2,t%p3,m%p4"; } [(set_attr "type" "vsetvl") (set_attr "mode" "<MODE>") (set (attr "sew") (symbol_ref "INTVAL (operands[1])")) @@ -3680,7 +3711,7 @@ (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3701,7 +3732,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3722,7 +3753,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c index 2e0e12aa045..2eef9e1e1a8 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c @@ -1,4 +1,4 @@ -/* { dg-do compile } */ +/* { dg-do compile { target { ! riscv_xtheadvector } } } */ /* { dg-skip-if "test rvv intrinsic" { *-*-* } { "*" } { "-march=rv*v*" } } */ void foo0 () {__rvv_bool64_t t;} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c index 3d81b179235..ef329e30785 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c @@ -1,4 +1,4 @@ /* { dg-do compile } */ /* { dg-options "-O3 -march=rv32gc -mabi=ilp32d" } */ -#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' extension enabled} } */ +#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' or 'XTHEADVECTOR' extension enabled} } */ diff --git a/gcc/testsuite/lib/target-supports.exp b/gcc/testsuite/lib/target-supports.exp index 7f13ff0ca56..70df6b1401c 100644 --- a/gcc/testsuite/lib/target-supports.exp +++ b/gcc/testsuite/lib/target-supports.exp @@ -1952,6 +1952,18 @@ proc check_effective_target_riscv_zbb { } { }] } +# Return 1 if the target arch supports the XTheadVector extension, 0 otherwise. +# Cache the result. + +proc check_effective_target_riscv_xtheadvector { } { + return [check_no_compiler_messages riscv_ext_xtheadvector assembly { + #ifndef __riscv_xtheadvector + #error "Not __riscv_xtheadvector" + #endif + }] +} + + # Return 1 if we can execute code when using dg-add-options riscv_v proc check_effective_target_riscv_v_ok { } { -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* [PATCH v4] RISC-V: Handle differences between XTheadvector and Vector 2024-01-03 6:15 ` Jun Sha (Joshua) @ 2024-01-04 2:29 ` Jun Sha (Joshua) 2024-01-09 3:18 ` [PATCH v5] " Jun Sha (Joshua) 2024-01-08 23:08 ` [PATCH v4] " 钟居哲 1 sibling, 1 reply; 133+ messages in thread From: Jun Sha (Joshua) @ 2024-01-04 2:29 UTC (permalink / raw) To: gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, Jun Sha (Joshua), Jin Ma, Xianmiao Qu This patch is to handle the differences in instruction generation between Vector and XTheadVector. In this version, we only support partial xtheadvector instructions that leverage directly from current RVV1.0 with simple adding "th." prefix. For different name xtheadvector instructions but share same patterns as RVV1.0 instructions, we will use ASM targethook to rewrite the whole string of the instructions in the following patches. For some vector patterns that cannot be avoided, we use "!TARGET_XTHEADVECTOR" to disable them in vector.md in order not to generate instructions that xtheadvector does not support, like vmv1r and vsext.vf2. gcc/ChangeLog: * config.gcc: Add files for XTheadVector intrinsics. * config/riscv/autovec.md: Guard XTheadVector. * config/riscv/riscv-c.cc: Add pragma for XTheadVector. * config/riscv/riscv-string.cc (expand_block_move): Guard XTheadVector. (get_prefer_tail_policy): Give specific value for tail. (get_prefer_mask_policy): Give specific value for mask. (vls_mode_valid_p): Avoid autovec. * config/riscv/riscv-vector-builtins-shapes.cc (check_type): (build_one): New function. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): (DEF_THEAD_RVV_FUNCTION): Add new marcos. (check_required_extensions): (handle_pragma_vector): * config/riscv/riscv-vector-builtins.h (RVV_REQUIRE_VECTOR): (RVV_REQUIRE_XTHEADVECTOR): Add RVV_REQUIRE_VECTOR and RVV_REQUIRE_XTHEADVECTOR. (struct function_group_info): * config/riscv/riscv-vector-switch.def (ENTRY): Disable fractional mode for the XTheadVector extension. (TUPLE_ENTRY): Likewise. * config/riscv/riscv-vsetvl.cc: Add functions for xtheadvector. * config/riscv/riscv.cc (riscv_v_ext_vls_mode_p): Guard XTheadVector. (riscv_v_adjust_bytesize): Likewise. (riscv_preferred_simd_mode): Likewsie. (riscv_autovectorize_vector_modes): Likewise. (riscv_vector_mode_supported_any_target_p): Likewise. (TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P): Likewise. * config/riscv/vector-iterators.md: Remove fractional LMUL. * config/riscv/vector.md: Include thead-vector.md. * config/riscv/riscv_th_vector.h: New file. * config/riscv/thead-vector.md: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/base/pragma-1.c: Add XTheadVector. * gcc.target/riscv/rvv/base/abi-1.c: Exclude XTheadVector. * lib/target-supports.exp: Add target for XTheadVector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config.gcc | 2 +- gcc/config/riscv/autovec.md | 2 +- gcc/config/riscv/predicates.md | 4 +- gcc/config/riscv/riscv-c.cc | 3 +- gcc/config/riscv/riscv-string.cc | 3 +- gcc/config/riscv/riscv-v.cc | 6 +- .../riscv/riscv-vector-builtins-bases.cc | 48 +++-- .../riscv/riscv-vector-builtins-shapes.cc | 23 +++ gcc/config/riscv/riscv-vector-switch.def | 150 +++++++------- gcc/config/riscv/riscv.cc | 20 +- gcc/config/riscv/riscv_th_vector.h | 49 +++++ gcc/config/riscv/thead-vector.md | 102 ++++++++++ gcc/config/riscv/vector-iterators.md | 186 +++++++++--------- gcc/config/riscv/vector.md | 43 +++- .../gcc.target/riscv/rvv/base/abi-1.c | 2 +- .../gcc.target/riscv/rvv/base/pragma-1.c | 2 +- gcc/testsuite/lib/target-supports.exp | 12 ++ 17 files changed, 451 insertions(+), 206 deletions(-) create mode 100644 gcc/config/riscv/riscv_th_vector.h create mode 100644 gcc/config/riscv/thead-vector.md diff --git a/gcc/config.gcc b/gcc/config.gcc index f0676c830e8..1445d98c147 100644 --- a/gcc/config.gcc +++ b/gcc/config.gcc @@ -549,7 +549,7 @@ riscv*) extra_objs="${extra_objs} riscv-vector-builtins.o riscv-vector-builtins-shapes.o riscv-vector-builtins-bases.o" extra_objs="${extra_objs} thead.o riscv-target-attr.o" d_target_objs="riscv-d.o" - extra_headers="riscv_vector.h" + extra_headers="riscv_vector.h riscv_th_vector.h" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.cc" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.h" ;; diff --git a/gcc/config/riscv/autovec.md b/gcc/config/riscv/autovec.md index 8b8a92f10a1..1fac56c7095 100644 --- a/gcc/config/riscv/autovec.md +++ b/gcc/config/riscv/autovec.md @@ -2579,7 +2579,7 @@ [(match_operand 0 "register_operand") (match_operand 1 "memory_operand") (match_operand:ANYI 2 "const_int_operand")] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" { riscv_vector::expand_rawmemchr(<MODE>mode, operands[0], operands[1], operands[2]); diff --git a/gcc/config/riscv/predicates.md b/gcc/config/riscv/predicates.md index b87a6900841..6e36c5afae1 100644 --- a/gcc/config/riscv/predicates.md +++ b/gcc/config/riscv/predicates.md @@ -428,7 +428,9 @@ ;; Predicates for the V extension. (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (ior (match_test "TARGET_XTHEADVECTOR && rtx_equal_p (op, const0_rtx)") + (match_test "!TARGET_XTHEADVECTOR")) + (match_operand 0 "const_csr_operand")))) (define_special_predicate "autovec_length_operand" (ior (match_operand 0 "pmode_register_operand") diff --git a/gcc/config/riscv/riscv-c.cc b/gcc/config/riscv/riscv-c.cc index d290c58613a..d6fb83ac00c 100644 --- a/gcc/config/riscv/riscv-c.cc +++ b/gcc/config/riscv/riscv-c.cc @@ -195,7 +195,8 @@ riscv_pragma_intrinsic (cpp_reader *) const char *name = TREE_STRING_POINTER (x); - if (strcmp (name, "vector") == 0) + if (strcmp (name, "vector") == 0 + || strcmp (name, "xtheadvector") == 0) { if (!TARGET_VECTOR) { diff --git a/gcc/config/riscv/riscv-string.cc b/gcc/config/riscv/riscv-string.cc index 11c1f74d0b3..f431b562222 100644 --- a/gcc/config/riscv/riscv-string.cc +++ b/gcc/config/riscv/riscv-string.cc @@ -773,7 +773,8 @@ riscv_expand_block_move_scalar (rtx dest, rtx src, rtx length) bool riscv_expand_block_move (rtx dest, rtx src, rtx length) { - if (TARGET_VECTOR && stringop_strategy & STRATEGY_VECTOR) + if ((TARGET_VECTOR && !TARGET_XTHEADVECTOR) + && stringop_strategy & STRATEGY_VECTOR) { bool ok = riscv_vector::expand_block_move (dest, src, length); if (ok) diff --git a/gcc/config/riscv/riscv-v.cc b/gcc/config/riscv/riscv-v.cc index 3c83be35715..22520e0bc29 100644 --- a/gcc/config/riscv/riscv-v.cc +++ b/gcc/config/riscv/riscv-v.cc @@ -1781,7 +1781,7 @@ get_prefer_tail_policy () compiler pick up either agnostic or undisturbed. Maybe we will have a compile option like -mprefer=agnostic to set this value???. */ - return TAIL_ANY; + return TARGET_XTHEADVECTOR ? TAIL_AGNOSTIC : TAIL_ANY; } /* Get prefer mask policy. */ @@ -1792,7 +1792,7 @@ get_prefer_mask_policy () compiler pick up either agnostic or undisturbed. Maybe we will have a compile option like -mprefer=agnostic to set this value???. */ - return MASK_ANY; + return TARGET_XTHEADVECTOR ? MASK_UNDISTURBED : MASK_ANY; } /* Get avl_type rtx. */ @@ -4397,7 +4397,7 @@ cmp_lmul_gt_one (machine_mode mode) bool vls_mode_valid_p (machine_mode vls_mode) { - if (!TARGET_VECTOR) + if (!TARGET_VECTOR || TARGET_XTHEADVECTOR) return false; if (riscv_autovec_preference == RVV_SCALABLE) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index c51affde353..22d877af9be 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -115,23 +115,37 @@ public: tree type = builtin_types[e.type.index].vector; machine_mode mode = TYPE_MODE (type); - /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. - - - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 - - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 - - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 - - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 - - e8,m2/e16,m4/e32,m8 --> e8m2 - - e8,m4/e16,m8 --> e8m4 - - e8,m8 --> e8m8 - */ - /* SEW. */ - e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); - - /* LMUL. */ - machine_mode e8_mode - = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); - e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + + if (TARGET_XTHEADVECTOR) + { + machine_mode inner_mode = GET_MODE_INNER (mode); + /* SEW. */ + e.add_input_operand (Pmode, + gen_int_mode (GET_MODE_BITSIZE (inner_mode), Pmode)); + /* LMUL. */ + e.add_input_operand (Pmode, + gen_int_mode (get_vlmul (mode), Pmode)); + } + else + { + /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. + + - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 + - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 + - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 + - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 + - e8,m2/e16,m4/e32,m8 --> e8m2 + - e8,m4/e16,m8 --> e8m4 + - e8,m8 --> e8m8 + */ + /* SEW. */ + e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); + + /* LMUL. */ + machine_mode e8_mode + = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); + e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + } /* TAIL_ANY. */ e.add_input_operand (Pmode, diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 4a754e0228f..6b49404a1fa 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -33,6 +33,25 @@ namespace riscv_vector { +/* Check whether the RETURN_TYPE and ARGUMENT_TYPES are + valid for the function. */ + +static bool +check_type (tree return_type, vec<tree> &argument_types) +{ + tree arg; + unsigned i; + + if (!return_type) + return false; + + FOR_EACH_VEC_ELT (argument_types, i, arg) + if (!arg) + return false; + + return true; +} + /* Add one function instance for GROUP, using operand suffix at index OI, mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ static void @@ -49,6 +68,10 @@ build_one (function_builder &b, const function_group_info &group, group.ops_infos.types[vec_type_idx].index); b.allocate_argument_types (function_instance, argument_types); b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + b.add_overloaded_function (function_instance, *group.shape); b.add_unique_function (function_instance, (*group.shape), return_type, argument_types); diff --git a/gcc/config/riscv/riscv-vector-switch.def b/gcc/config/riscv/riscv-vector-switch.def index 5c9f9bcbc3e..f7a66b34bae 100644 --- a/gcc/config/riscv/riscv-vector-switch.def +++ b/gcc/config/riscv/riscv-vector-switch.def @@ -68,9 +68,9 @@ Encode the ratio of SEW/LMUL into the mask types. #endif /* Disable modes if TARGET_MIN_VLEN == 32. */ -ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) -ENTRY (RVVMF32BI, true, LMUL_F4, 32) -ENTRY (RVVMF16BI, true, LMUL_F2, 16) +ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F8, 64) +ENTRY (RVVMF32BI, true, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F4, 32) +ENTRY (RVVMF16BI, true, TARGET_XTHEADVECTOR ? LMUL_1 : LMUL_F2 , 16) ENTRY (RVVMF8BI, true, LMUL_1, 8) ENTRY (RVVMF4BI, true, LMUL_2, 4) ENTRY (RVVMF2BI, true, LMUL_4, 2) @@ -81,39 +81,39 @@ ENTRY (RVVM8QI, true, LMUL_8, 1) ENTRY (RVVM4QI, true, LMUL_4, 2) ENTRY (RVVM2QI, true, LMUL_2, 4) ENTRY (RVVM1QI, true, LMUL_1, 8) -ENTRY (RVVMF2QI, true, LMUL_F2, 16) -ENTRY (RVVMF4QI, true, LMUL_F4, 32) -ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) +ENTRY (RVVMF2QI, !TARGET_XTHEADVECTOR, LMUL_F2, 16) +ENTRY (RVVMF4QI, !TARGET_XTHEADVECTOR, LMUL_F4, 32) +ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F8, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8HI, true, LMUL_8, 2) ENTRY (RVVM4HI, true, LMUL_4, 4) ENTRY (RVVM2HI, true, LMUL_2, 8) ENTRY (RVVM1HI, true, LMUL_1, 16) -ENTRY (RVVMF2HI, true, LMUL_F2, 32) -ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HI, !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_16. */ ENTRY (RVVM8HF, TARGET_VECTOR_ELEN_FP_16, LMUL_8, 2) ENTRY (RVVM4HF, TARGET_VECTOR_ELEN_FP_16, LMUL_4, 4) ENTRY (RVVM2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_2, 8) ENTRY (RVVM1HF, TARGET_VECTOR_ELEN_FP_16, LMUL_1, 16) -ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_F2, 32) -ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8SI, true, LMUL_8, 4) ENTRY (RVVM4SI, true, LMUL_4, 8) ENTRY (RVVM2SI, true, LMUL_2, 16) ENTRY (RVVM1SI, true, LMUL_1, 32) -ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_32. */ ENTRY (RVVM8SF, TARGET_VECTOR_ELEN_FP_32, LMUL_8, 4) ENTRY (RVVM4SF, TARGET_VECTOR_ELEN_FP_32, LMUL_4, 8) ENTRY (RVVM2SF, TARGET_VECTOR_ELEN_FP_32, LMUL_2, 16) ENTRY (RVVM1SF, TARGET_VECTOR_ELEN_FP_32, LMUL_1, 32) -ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if !TARGET_VECTOR_ELEN_64. */ ENTRY (RVVM8DI, TARGET_VECTOR_ELEN_64, LMUL_8, 8) @@ -140,127 +140,127 @@ ENTRY (RVVM1DF, TARGET_VECTOR_ELEN_FP_64, LMUL_1, 64) #endif TUPLE_ENTRY (RVVM1x8QI, true, RVVM1QI, 8, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x8QI, true, RVVMF2QI, 8, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x8QI, true, RVVMF4QI, 8, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 8, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x8QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 8, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x8QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 8, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 8, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x7QI, true, RVVM1QI, 7, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x7QI, true, RVVMF2QI, 7, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x7QI, true, RVVMF4QI, 7, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 7, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x7QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 7, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x7QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 7, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 7, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x6QI, true, RVVM1QI, 6, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x6QI, true, RVVMF2QI, 6, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x6QI, true, RVVMF4QI, 6, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 6, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x6QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 6, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x6QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 6, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 6, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x5QI, true, RVVM1QI, 5, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x5QI, true, RVVMF2QI, 5, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x5QI, true, RVVMF4QI, 5, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 5, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x5QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 5, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x5QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 5, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 5, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x4QI, true, RVVM2QI, 4, LMUL_2, 4) TUPLE_ENTRY (RVVM1x4QI, true, RVVM1QI, 4, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x4QI, true, RVVMF2QI, 4, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x4QI, true, RVVMF4QI, 4, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 4, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x4QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 4, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x4QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 4, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 4, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x3QI, true, RVVM2QI, 3, LMUL_2, 4) TUPLE_ENTRY (RVVM1x3QI, true, RVVM1QI, 3, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x3QI, true, RVVMF2QI, 3, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x3QI, true, RVVMF4QI, 3, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 3, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x3QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 3, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x3QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 3, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 3, LMUL_F8, 64) TUPLE_ENTRY (RVVM4x2QI, true, RVVM4QI, 2, LMUL_4, 2) TUPLE_ENTRY (RVVM2x2QI, true, RVVM2QI, 2, LMUL_2, 4) TUPLE_ENTRY (RVVM1x2QI, true, RVVM1QI, 2, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x2QI, true, RVVMF2QI, 2, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x2QI, true, RVVMF4QI, 2, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 2, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x2QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 2, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x2QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 2, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 2, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x8HI, true, RVVM1HI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HI, true, RVVMF2HI, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HI, true, RVVM1HI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HI, true, RVVMF2HI, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HI, true, RVVM1HI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HI, true, RVVMF2HI, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HI, true, RVVM1HI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HI, true, RVVMF2HI, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HI, true, RVVM2HI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HI, true, RVVM1HI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HI, true, RVVMF2HI, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HI, true, RVVM2HI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HI, true, RVVM1HI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HI, true, RVVMF2HI, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HI, true, RVVM4HI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HI, true, RVVM2HI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HI, true, RVVM1HI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HI, true, RVVMF2HI, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM4HF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8SI, true, RVVM1SI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SI, true, RVVM1SI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SI, true, RVVM1SI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SI, true, RVVM1SI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SI, true, RVVM2SI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SI, true, RVVM1SI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SI, true, RVVM2SI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SI, true, RVVM1SI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SI, true, RVVM4SI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SI, true, RVVM2SI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SI, true, RVVM1SI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM4SF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 8, LMUL_1, 16) TUPLE_ENTRY (RVVM1x7DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 7, LMUL_1, 16) diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index 51878797287..44a36db7803 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -1406,6 +1406,9 @@ riscv_v_adjust_bytesize (machine_mode mode, int scale) { if (riscv_v_ext_vector_mode_p (mode)) { + if (TARGET_XTHEADVECTOR) + return BYTES_PER_RISCV_VECTOR; + poly_int64 nunits = GET_MODE_NUNITS (mode); poly_int64 mode_size = GET_MODE_SIZE (mode); @@ -9962,7 +9965,7 @@ riscv_use_divmod_expander (void) static machine_mode riscv_preferred_simd_mode (scalar_mode mode) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::preferred_simd_mode (mode); return word_mode; @@ -10313,7 +10316,7 @@ riscv_mode_priority (int, int n) unsigned int riscv_autovectorize_vector_modes (vector_modes *modes, bool all) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::autovectorize_vector_modes (modes, all); return default_autovectorize_vector_modes (modes, all); @@ -10496,6 +10499,16 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) return false; } +/* Implements target hook vector_mode_supported_any_target_p. */ + +static bool +riscv_vector_mode_supported_any_target_p (machine_mode mode) +{ + if (TARGET_XTHEADVECTOR) + return false; + return true; +} + /* Initialize the GCC target structure. */ #undef TARGET_ASM_ALIGNED_HI_OP #define TARGET_ASM_ALIGNED_HI_OP "\t.half\t" @@ -10839,6 +10852,9 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) #undef TARGET_PREFERRED_ELSE_VALUE #define TARGET_PREFERRED_ELSE_VALUE riscv_preferred_else_value +#undef TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P +#define TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P riscv_vector_mode_supported_any_target_p + struct gcc_target targetm = TARGET_INITIALIZER; #include "gt-riscv.h" diff --git a/gcc/config/riscv/riscv_th_vector.h b/gcc/config/riscv/riscv_th_vector.h new file mode 100644 index 00000000000..a897b9cb90a --- /dev/null +++ b/gcc/config/riscv/riscv_th_vector.h @@ -0,0 +1,49 @@ +/* RISC-V 'XTheadVector' Extension intrinsics include file. + Copyright (C) 2022-2023 Free Software Foundation, Inc. + + This file is part of GCC. + + GCC is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published + by the Free Software Foundation; either version 3, or (at your + option) any later version. + + GCC is distributed in the hope that it will be useful, but WITHOUT + ANY WARRANTY; without even the implied warranty of MERCHANTABILITY + or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public + License for more details. + + Under Section 7 of GPL version 3, you are granted additional + permissions described in the GCC Runtime Library Exception, version + 3.1, as published by the Free Software Foundation. + + You should have received a copy of the GNU General Public License and + a copy of the GCC Runtime Library Exception along with this program; + see the files COPYING3 and COPYING.RUNTIME respectively. If not, see + <http://www.gnu.org/licenses/>. */ + +#ifndef __RISCV_TH_VECTOR_H +#define __RISCV_TH_VECTOR_H + +#include <stdint.h> +#include <stddef.h> + +#ifndef __riscv_xtheadvector +#error "XTheadVector intrinsics require the xtheadvector extension." +#else +#ifdef __cplusplus +extern "C" { +#endif + +/* NOTE: This implementation of riscv_th_vector.h is intentionally short. It does + not define the RVV types and intrinsic functions directly in C and C++ + code, but instead uses the following pragma to tell GCC to insert the + necessary type and function definitions itself. The net effect is the + same, and the file is a complete implementation of riscv_th_vector.h. */ +#pragma riscv intrinsic "xtheadvector" + +#ifdef __cplusplus +} +#endif // __cplusplus +#endif // __riscv_xtheadvector +#endif // __RISCV_TH_ECTOR_H diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md new file mode 100644 index 00000000000..696b815252d --- /dev/null +++ b/gcc/config/riscv/thead-vector.md @@ -0,0 +1,102 @@ +(define_c_enum "unspec" [ + UNSPEC_TH_VWLDST +]) + +(define_mode_iterator V_VLS_VT [V VLS VT]) +(define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) + +(define_split + [(set (match_operand:V_VB_VLS_VT 0 "reg_or_mem_operand") + (match_operand:V_VB_VLS_VT 1 "reg_or_mem_operand"))] + "TARGET_XTHEADVECTOR" + [(const_int 0)] + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + }) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:V_VLS_VT 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:V_VLS_VT + [(match_operand:V_VLS_VT 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2")]) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:VB 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:VB + [(match_operand:VB 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2") + (set (attr "sew") (const_int 8)) + (set (attr "vlmul") (symbol_ref "riscv_vector::LMUL_1"))]) + +(define_insn_and_split "*pred_th_mov<mode>" + [(set (match_operand:VB_VLS 0 "nonimmediate_operand" "=vr, m, vr, vr, vr") + (if_then_else:VB_VLS + (unspec:VB_VLS + [(match_operand:VB_VLS 1 "vector_all_trues_mask_operand" "Wc1, Wc1, Wc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) + (match_operand:VB_VLS 3 "vector_move_operand" " m, vr, vr, Wc0, Wc1") + (match_operand:VB_VLS 2 "vector_undef_operand" " vu, vu, vu, vu, vu")))] + "TARGET_XTHEADVECTOR" + "@ + # + # + vmcpy.m\t%0,%3 + vmclr.m\t%0 + vmset.m\t%0" + "&& !reload_completed" + [(const_int 0)] + { + if ((MEM_P (operands[0]) || MEM_P (operands[3])) + || (REG_P (operands[0]) && REG_P (operands[3]) + && INTVAL (operands[5]) == riscv_vector::VLMAX)) + { + emit_move_insn (operands[0], operands[3]); + DONE; + } + FAIL; + } + [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/config/riscv/vector-iterators.md b/gcc/config/riscv/vector-iterators.md index 317dc9de253..c564c4d06e1 100644 --- a/gcc/config/riscv/vector-iterators.md +++ b/gcc/config/riscv/vector-iterators.md @@ -109,11 +109,11 @@ ]) (define_mode_iterator VI [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -128,11 +128,11 @@ ;; allow the instruction and mode to be matched during combine et al. (define_mode_iterator VF [ (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -140,11 +140,11 @@ (define_mode_iterator VF_ZVFHMIN [ (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -271,16 +271,16 @@ ]) (define_mode_iterator VEEWEXT2 [ - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -290,10 +290,10 @@ ]) (define_mode_iterator VEEWEXT4 [ - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -311,59 +311,59 @@ ]) (define_mode_iterator VEEWTRUNC2 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") (RVVM4SI "TARGET_64BIT") (RVVM2SI "TARGET_64BIT") (RVVM1SI "TARGET_64BIT") - (RVVMF2SI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEEWTRUNC4 [ - RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM2HI "TARGET_64BIT") (RVVM1HI "TARGET_64BIT") - (RVVMF2HI "TARGET_64BIT") - (RVVMF4HI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2HI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") (RVVM1HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEEWTRUNC8 [ (RVVM1QI "TARGET_64BIT") - (RVVMF2QI "TARGET_64BIT") - (RVVMF4QI "TARGET_64BIT") - (RVVMF8QI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2QI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF4QI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEI16 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -452,11 +452,11 @@ ]) (define_mode_iterator VFULLI [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_FULL_V") (RVVM4DI "TARGET_FULL_V") (RVVM2DI "TARGET_FULL_V") (RVVM1DI "TARGET_FULL_V") @@ -509,17 +509,17 @@ ]) (define_mode_iterator VI_QH [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") ]) (define_mode_iterator VI_QHS [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -560,11 +560,11 @@ ]) (define_mode_iterator VI_QHS_NO_M8 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -603,11 +603,11 @@ (define_mode_iterator VF_HS [ (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (V1HF "riscv_vector::vls_mode_valid_p (V1HFmode) && TARGET_ZVFH") (V2HF "riscv_vector::vls_mode_valid_p (V2HFmode) && TARGET_ZVFH") @@ -638,12 +638,12 @@ (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") - (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (V1HF "riscv_vector::vls_mode_valid_p (V1HFmode) && TARGET_ZVFH") (V2HF "riscv_vector::vls_mode_valid_p (V2HFmode) && TARGET_ZVFH") @@ -674,11 +674,11 @@ ]) (define_mode_iterator V_VLSI_QHS [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -756,27 +756,27 @@ ;; E.g. when index mode = RVVM8QImde and Pmode = SImode, if it is not zero_extend or ;; scalar != 1, such gather/scatter is not allowed since we don't have RVVM32SImode. (define_mode_iterator RATIO64 [ - (RVVMF8QI "TARGET_MIN_VLEN > 32") - (RVVMF4HI "TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM1DI "TARGET_VECTOR_ELEN_64") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") ]) (define_mode_iterator RATIO32 [ - RVVMF4QI - RVVMF2HI + (RVVMF4QI "!TARGET_XTHEADVECTOR") + (RVVMF2HI "!TARGET_XTHEADVECTOR") RVVM1SI (RVVM2DI "TARGET_VECTOR_ELEN_64") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") ]) (define_mode_iterator RATIO16 [ - RVVMF2QI + (RVVMF2QI "!TARGET_XTHEADVECTOR") RVVM1HI RVVM2SI (RVVM4DI "TARGET_VECTOR_ELEN_64") @@ -814,21 +814,21 @@ ]) (define_mode_iterator RATIO64I [ - (RVVMF8QI "TARGET_MIN_VLEN > 32") - (RVVMF4HI "TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM1DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") ]) (define_mode_iterator RATIO32I [ - RVVMF4QI - RVVMF2HI + (RVVMF4QI "!TARGET_XTHEADVECTOR") + (RVVMF2HI "!TARGET_XTHEADVECTOR") RVVM1SI (RVVM2DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") ]) (define_mode_iterator RATIO16I [ - RVVMF2QI + (RVVMF2QI "!TARGET_XTHEADVECTOR") RVVM1HI RVVM2SI (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") @@ -873,21 +873,21 @@ ]) (define_mode_iterator V_FRACT [ - RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") ]) (define_mode_iterator VWEXTI [ - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -933,7 +933,7 @@ (RVVM4SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -966,7 +966,7 @@ (RVVM4SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -996,7 +996,7 @@ (define_mode_iterator VWCONVERTI [ (RVVM8SI "TARGET_ZVFH") (RVVM4SI "TARGET_ZVFH") (RVVM2SI "TARGET_ZVFH") (RVVM1SI "TARGET_ZVFH") - (RVVMF2SI "TARGET_ZVFH") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_ZVFH") (RVVM8DI "TARGET_VECTOR_ELEN_64 && TARGET_VECTOR_ELEN_FP_32") (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_VECTOR_ELEN_FP_32") @@ -1045,7 +1045,7 @@ ]) (define_mode_iterator VQEXTI [ - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -1456,11 +1456,11 @@ ;; VINDEXED [VI8 VI16 VI32 (VI64 "TARGET_64BIT")]. (define_mode_iterator VINDEXED [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") @@ -1468,12 +1468,12 @@ (RVVM1DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64 && TARGET_64BIT") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64 && TARGET_64BIT") @@ -3173,11 +3173,11 @@ (define_mode_iterator V_VLS_F_CONVERT_SI [ (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") - (RVVMF2HF "TARGET_ZVFH") (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") @@ -3290,12 +3290,12 @@ ]) (define_mode_iterator V_VLS_F_CONVERT_DI [ - (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index 24f91f058ef..2a5ab487b1c 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -87,7 +87,7 @@ ;; check. However, we need default value of SEW for vsetvl instruction since there ;; is no field for ratio in the vsetvl instruction encoding. (define_attr "sew" "" - (cond [(eq_attr "mode" "RVVMF64BI,RVVMF32BI,RVVMF16BI,RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ + (cond [(eq_attr "mode" "RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ RVVM8QI,RVVM4QI,RVVM2QI,RVVM1QI,RVVMF2QI,RVVMF4QI,RVVMF8QI,\ RVVM1x8QI,RVVMF2x8QI,RVVMF4x8QI,RVVMF8x8QI,\ RVVM1x7QI,RVVMF2x7QI,RVVMF4x7QI,RVVMF8x7QI,\ @@ -99,6 +99,18 @@ V1QI,V2QI,V4QI,V8QI,V16QI,V32QI,V64QI,V128QI,V256QI,V512QI,V1024QI,V2048QI,V4096QI,\ V1BI,V2BI,V4BI,V8BI,V16BI,V32BI,V64BI,V128BI,V256BI,V512BI,V1024BI,V2048BI,V4096BI") (const_int 8) + (eq_attr "mode" "RVVMF16BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 16) + (const_int 8)) + (eq_attr "mode" "RVVMF32BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 32) + (const_int 8)) + (eq_attr "mode" "RVVMF64BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 64) + (const_int 8)) (eq_attr "mode" "RVVM8HI,RVVM4HI,RVVM2HI,RVVM1HI,RVVMF2HI,RVVMF4HI,\ RVVM1x8HI,RVVMF2x8HI,RVVMF4x8HI,\ RVVM1x7HI,RVVMF2x7HI,RVVMF4x7HI,\ @@ -1078,6 +1090,12 @@ - We can not leave it to TARGET_SECONDARY_RELOAD since it happens before spilling. The clobber scratch is used by spilling fractional registers in IRA/LRA so it's too early. */ + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; @@ -1114,7 +1132,7 @@ (define_insn "*mov<mode>_whole" [(set (match_operand:V_WHOLE 0 "reg_or_mem_operand" "=vr, m,vr") (match_operand:V_WHOLE 1 "reg_or_mem_operand" " m,vr,vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "@ vl%m1re<sew>.v\t%0,%1 vs%m1r.v\t%1,%0 @@ -1135,6 +1153,13 @@ (match_operand:VB 1 "general_operand"))] "TARGET_VECTOR" { + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } + if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; }) @@ -1142,7 +1167,7 @@ (define_insn "*mov<mode>" [(set (match_operand:VB 0 "register_operand" "=vr") (match_operand:VB 1 "register_operand" " vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "vmv1r.v\t%0,%1" [(set_attr "type" "vmov") (set_attr "mode" "<MODE>")]) @@ -1528,7 +1553,7 @@ (match_dup 4) (match_dup 5)] UNSPEC_VSETVL))] "TARGET_VECTOR" - "vset%i1vli\t%0,%1,e%2,%m3,t%p4,m%p5" + { return TARGET_XTHEADVECTOR ? "vsetvli\t%0,%1,e%2,%m3" : "vset%i1vli\t%0,%1,e%2,%m3,t%p4,m%p5"; } [(set_attr "type" "vsetvl") (set_attr "mode" "<MODE>") (set (attr "sew") (symbol_ref "INTVAL (operands[2])")) @@ -1546,7 +1571,7 @@ (match_operand 2 "const_int_operand" "i") (match_operand 3 "const_int_operand" "i")] UNSPEC_VSETVL))] "TARGET_VECTOR" - "vsetvli\tzero,zero,e%0,%m1,t%p2,m%p3" + { return TARGET_XTHEADVECTOR ? "vsetvli\tzero,zero,e%0,%m1" : "vsetvli\tzero,zero,e%0,%m1,t%p2,m%p3"; } [(set_attr "type" "vsetvl") (set_attr "mode" "SI") (set (attr "sew") (symbol_ref "INTVAL (operands[0])")) @@ -1568,7 +1593,7 @@ (match_operand 3 "const_int_operand" "i") (match_operand 4 "const_int_operand" "i")] UNSPEC_VSETVL))] "TARGET_VECTOR" - "vset%i0vli\tzero,%0,e%1,%m2,t%p3,m%p4" + { return TARGET_XTHEADVECTOR ? "vsetvli\tzero,%0,e%1,%m2" : "vset%i0vli\tzero,%0,e%1,%m2,t%p3,m%p4"; } [(set_attr "type" "vsetvl") (set_attr "mode" "<MODE>") (set (attr "sew") (symbol_ref "INTVAL (operands[1])")) @@ -3697,7 +3722,7 @@ (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3718,7 +3743,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3739,7 +3764,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c index 2e0e12aa045..2eef9e1e1a8 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c @@ -1,4 +1,4 @@ -/* { dg-do compile } */ +/* { dg-do compile { target { ! riscv_xtheadvector } } } */ /* { dg-skip-if "test rvv intrinsic" { *-*-* } { "*" } { "-march=rv*v*" } } */ void foo0 () {__rvv_bool64_t t;} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c index 3d81b179235..ef329e30785 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c @@ -1,4 +1,4 @@ /* { dg-do compile } */ /* { dg-options "-O3 -march=rv32gc -mabi=ilp32d" } */ -#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' extension enabled} } */ +#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' or 'XTHEADVECTOR' extension enabled} } */ diff --git a/gcc/testsuite/lib/target-supports.exp b/gcc/testsuite/lib/target-supports.exp index 3aa761d3e09..a4205e8eb4f 100644 --- a/gcc/testsuite/lib/target-supports.exp +++ b/gcc/testsuite/lib/target-supports.exp @@ -1956,6 +1956,18 @@ proc check_effective_target_riscv_zbb { } { }] } +# Return 1 if the target arch supports the XTheadVector extension, 0 otherwise. +# Cache the result. + +proc check_effective_target_riscv_xtheadvector { } { + return [check_no_compiler_messages riscv_ext_xtheadvector assembly { + #ifndef __riscv_xtheadvector + #error "Not __riscv_xtheadvector" + #endif + }] +} + + # Return 1 if we can execute code when using dg-add-options riscv_v proc check_effective_target_riscv_v_ok { } { -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* [PATCH v5] RISC-V: Handle differences between XTheadvector and Vector 2024-01-04 2:29 ` Jun Sha (Joshua) @ 2024-01-09 3:18 ` Jun Sha (Joshua) 2024-01-09 22:33 ` 钟居哲 2024-01-10 2:22 ` Jun Sha (Joshua) 0 siblings, 2 replies; 133+ messages in thread From: Jun Sha (Joshua) @ 2024-01-09 3:18 UTC (permalink / raw) To: gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, Jun Sha (Joshua), Jin Ma, Xianmiao Qu This patch is to handle the differences in instruction generation between Vector and XTheadVector. In this version, we only support partial xtheadvector instructions that leverage directly from current RVV1.0 with simple adding "th." prefix. For different name xtheadvector instructions but share same patterns as RVV1.0 instructions, we will use ASM targethook to rewrite the whole string of the instructions in the following patches. For some vector patterns that cannot be avoided, we use "!TARGET_XTHEADVECTOR" to disable them in vector.md in order not to generate instructions that xtheadvector does not support, like vmv1r and vsext.vf2. gcc/ChangeLog: * config.gcc: Add files for XTheadVector intrinsics. * config/riscv/autovec.md: Guard XTheadVector. * config/riscv/riscv-c.cc: Add pragma for XTheadVector. * config/riscv/riscv-string.cc (expand_block_move): Guard XTheadVector. (get_prefer_tail_policy): Give specific value for tail. (get_prefer_mask_policy): Give specific value for mask. (vls_mode_valid_p): Avoid autovec. * config/riscv/riscv-vector-builtins-shapes.cc (check_type): (build_one): New function. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): (DEF_THEAD_RVV_FUNCTION): Add new marcos. (check_required_extensions): (handle_pragma_vector): * config/riscv/riscv-vector-builtins.h (RVV_REQUIRE_VECTOR): (RVV_REQUIRE_XTHEADVECTOR): Add RVV_REQUIRE_VECTOR and RVV_REQUIRE_XTHEADVECTOR. (struct function_group_info): * config/riscv/riscv-vector-switch.def (ENTRY): Disable fractional mode for the XTheadVector extension. (TUPLE_ENTRY): Likewise. * config/riscv/riscv-vsetvl.cc: Add functions for xtheadvector. * config/riscv/riscv.cc (riscv_v_ext_vls_mode_p): Guard XTheadVector. (riscv_v_adjust_bytesize): Likewise. (riscv_preferred_simd_mode): Likewsie. (riscv_autovectorize_vector_modes): Likewise. (riscv_vector_mode_supported_any_target_p): Likewise. (TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P): Likewise. * config/riscv/vector-iterators.md: Remove fractional LMUL. * config/riscv/vector.md: Include thead-vector.md. * config/riscv/riscv_th_vector.h: New file. * config/riscv/thead-vector.md: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/base/pragma-1.c: Add XTheadVector. * gcc.target/riscv/rvv/base/abi-1.c: Exclude XTheadVector. * lib/target-supports.exp: Add target for XTheadVector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config.gcc | 2 +- gcc/config/riscv/autovec.md | 2 +- gcc/config/riscv/predicates.md | 4 +- gcc/config/riscv/riscv-c.cc | 3 +- gcc/config/riscv/riscv-string.cc | 3 +- gcc/config/riscv/riscv-v.cc | 2 +- .../riscv/riscv-vector-builtins-bases.cc | 48 +++-- .../riscv/riscv-vector-builtins-shapes.cc | 23 +++ gcc/config/riscv/riscv-vector-switch.def | 150 +++++++------- gcc/config/riscv/riscv.cc | 20 +- gcc/config/riscv/riscv_th_vector.h | 49 +++++ gcc/config/riscv/thead-vector.md | 102 ++++++++++ gcc/config/riscv/thead.cc | 23 ++- gcc/config/riscv/vector-iterators.md | 186 +++++++++--------- gcc/config/riscv/vector.md | 49 ++++- .../gcc.target/riscv/rvv/base/abi-1.c | 2 +- .../gcc.target/riscv/rvv/base/pragma-1.c | 2 +- gcc/testsuite/lib/target-supports.exp | 12 ++ 18 files changed, 476 insertions(+), 206 deletions(-) create mode 100644 gcc/config/riscv/riscv_th_vector.h create mode 100644 gcc/config/riscv/thead-vector.md diff --git a/gcc/config.gcc b/gcc/config.gcc index 7e583390024..047e4c02cf4 100644 --- a/gcc/config.gcc +++ b/gcc/config.gcc @@ -549,7 +549,7 @@ riscv*) extra_objs="${extra_objs} riscv-vector-builtins.o riscv-vector-builtins-shapes.o riscv-vector-builtins-bases.o" extra_objs="${extra_objs} thead.o riscv-target-attr.o" d_target_objs="riscv-d.o" - extra_headers="riscv_vector.h" + extra_headers="riscv_vector.h riscv_th_vector.h" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.cc" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.h" ;; diff --git a/gcc/config/riscv/autovec.md b/gcc/config/riscv/autovec.md index 775eaa825b0..0477781cabe 100644 --- a/gcc/config/riscv/autovec.md +++ b/gcc/config/riscv/autovec.md @@ -2579,7 +2579,7 @@ [(match_operand 0 "register_operand") (match_operand 1 "memory_operand") (match_operand:ANYI 2 "const_int_operand")] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" { riscv_vector::expand_rawmemchr(<MODE>mode, operands[0], operands[1], operands[2]); diff --git a/gcc/config/riscv/predicates.md b/gcc/config/riscv/predicates.md index b1a79cae50a..0337da88284 100644 --- a/gcc/config/riscv/predicates.md +++ b/gcc/config/riscv/predicates.md @@ -428,7 +428,9 @@ ;; Predicates for the V extension. (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (ior (match_test "TARGET_XTHEADVECTOR && rtx_equal_p (op, const0_rtx)") + (match_test "!TARGET_XTHEADVECTOR")) + (match_operand 0 "const_csr_operand")))) (define_special_predicate "autovec_length_operand" (ior (match_operand 0 "pmode_register_operand") diff --git a/gcc/config/riscv/riscv-c.cc b/gcc/config/riscv/riscv-c.cc index 422ddc2c308..2e306057347 100644 --- a/gcc/config/riscv/riscv-c.cc +++ b/gcc/config/riscv/riscv-c.cc @@ -195,7 +195,8 @@ riscv_pragma_intrinsic (cpp_reader *) const char *name = TREE_STRING_POINTER (x); - if (strcmp (name, "vector") == 0) + if (strcmp (name, "vector") == 0 + || strcmp (name, "xtheadvector") == 0) { if (!TARGET_VECTOR) { diff --git a/gcc/config/riscv/riscv-string.cc b/gcc/config/riscv/riscv-string.cc index f227b8667ce..b09b51d7526 100644 --- a/gcc/config/riscv/riscv-string.cc +++ b/gcc/config/riscv/riscv-string.cc @@ -773,7 +773,8 @@ riscv_expand_block_move_scalar (rtx dest, rtx src, rtx length) bool riscv_expand_block_move (rtx dest, rtx src, rtx length) { - if (TARGET_VECTOR && stringop_strategy & STRATEGY_VECTOR) + if ((TARGET_VECTOR && !TARGET_XTHEADVECTOR) + && stringop_strategy & STRATEGY_VECTOR) { bool ok = riscv_vector::expand_block_move (dest, src, length); if (ok) diff --git a/gcc/config/riscv/riscv-v.cc b/gcc/config/riscv/riscv-v.cc index 2491522191a..7e25deb966d 100644 --- a/gcc/config/riscv/riscv-v.cc +++ b/gcc/config/riscv/riscv-v.cc @@ -4392,7 +4392,7 @@ cmp_lmul_gt_one (machine_mode mode) bool vls_mode_valid_p (machine_mode vls_mode) { - if (!TARGET_VECTOR) + if (!TARGET_VECTOR || TARGET_XTHEADVECTOR) return false; if (riscv_autovec_preference == RVV_SCALABLE) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 91af2c00674..46f1a1da33e 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -115,23 +115,37 @@ public: tree type = builtin_types[e.type.index].vector; machine_mode mode = TYPE_MODE (type); - /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. - - - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 - - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 - - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 - - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 - - e8,m2/e16,m4/e32,m8 --> e8m2 - - e8,m4/e16,m8 --> e8m4 - - e8,m8 --> e8m8 - */ - /* SEW. */ - e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); - - /* LMUL. */ - machine_mode e8_mode - = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); - e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + + if (TARGET_XTHEADVECTOR) + { + machine_mode inner_mode = GET_MODE_INNER (mode); + /* SEW. */ + e.add_input_operand (Pmode, + gen_int_mode (GET_MODE_BITSIZE (inner_mode), Pmode)); + /* LMUL. */ + e.add_input_operand (Pmode, + gen_int_mode (get_vlmul (mode), Pmode)); + } + else + { + /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. + + - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 + - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 + - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 + - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 + - e8,m2/e16,m4/e32,m8 --> e8m2 + - e8,m4/e16,m8 --> e8m4 + - e8,m8 --> e8m8 + */ + /* SEW. */ + e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); + + /* LMUL. */ + machine_mode e8_mode + = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); + e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + } /* TAIL_ANY. */ e.add_input_operand (Pmode, diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index ee8058dcdb1..1e4f4d53de6 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -33,6 +33,25 @@ namespace riscv_vector { +/* Check whether the RETURN_TYPE and ARGUMENT_TYPES are + valid for the function. */ + +static bool +check_type (tree return_type, vec<tree> &argument_types) +{ + tree arg; + unsigned i; + + if (!return_type) + return false; + + FOR_EACH_VEC_ELT (argument_types, i, arg) + if (!arg) + return false; + + return true; +} + /* Add one function instance for GROUP, using operand suffix at index OI, mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ static void @@ -49,6 +68,10 @@ build_one (function_builder &b, const function_group_info &group, group.ops_infos.types[vec_type_idx].index); b.allocate_argument_types (function_instance, argument_types); b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + b.add_overloaded_function (function_instance, *group.shape); b.add_unique_function (function_instance, (*group.shape), return_type, argument_types); diff --git a/gcc/config/riscv/riscv-vector-switch.def b/gcc/config/riscv/riscv-vector-switch.def index 1ad26c2a3b2..452283b7416 100644 --- a/gcc/config/riscv/riscv-vector-switch.def +++ b/gcc/config/riscv/riscv-vector-switch.def @@ -68,9 +68,9 @@ Encode the ratio of SEW/LMUL into the mask types. #endif /* Disable modes if TARGET_MIN_VLEN == 32. */ -ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) -ENTRY (RVVMF32BI, true, LMUL_F4, 32) -ENTRY (RVVMF16BI, true, LMUL_F2, 16) +ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F8, 64) +ENTRY (RVVMF32BI, true, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F4, 32) +ENTRY (RVVMF16BI, true, TARGET_XTHEADVECTOR ? LMUL_1 : LMUL_F2 , 16) ENTRY (RVVMF8BI, true, LMUL_1, 8) ENTRY (RVVMF4BI, true, LMUL_2, 4) ENTRY (RVVMF2BI, true, LMUL_4, 2) @@ -81,39 +81,39 @@ ENTRY (RVVM8QI, true, LMUL_8, 1) ENTRY (RVVM4QI, true, LMUL_4, 2) ENTRY (RVVM2QI, true, LMUL_2, 4) ENTRY (RVVM1QI, true, LMUL_1, 8) -ENTRY (RVVMF2QI, true, LMUL_F2, 16) -ENTRY (RVVMF4QI, true, LMUL_F4, 32) -ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) +ENTRY (RVVMF2QI, !TARGET_XTHEADVECTOR, LMUL_F2, 16) +ENTRY (RVVMF4QI, !TARGET_XTHEADVECTOR, LMUL_F4, 32) +ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F8, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8HI, true, LMUL_8, 2) ENTRY (RVVM4HI, true, LMUL_4, 4) ENTRY (RVVM2HI, true, LMUL_2, 8) ENTRY (RVVM1HI, true, LMUL_1, 16) -ENTRY (RVVMF2HI, true, LMUL_F2, 32) -ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HI, !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_16. */ ENTRY (RVVM8HF, TARGET_VECTOR_ELEN_FP_16, LMUL_8, 2) ENTRY (RVVM4HF, TARGET_VECTOR_ELEN_FP_16, LMUL_4, 4) ENTRY (RVVM2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_2, 8) ENTRY (RVVM1HF, TARGET_VECTOR_ELEN_FP_16, LMUL_1, 16) -ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_F2, 32) -ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8SI, true, LMUL_8, 4) ENTRY (RVVM4SI, true, LMUL_4, 8) ENTRY (RVVM2SI, true, LMUL_2, 16) ENTRY (RVVM1SI, true, LMUL_1, 32) -ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_32. */ ENTRY (RVVM8SF, TARGET_VECTOR_ELEN_FP_32, LMUL_8, 4) ENTRY (RVVM4SF, TARGET_VECTOR_ELEN_FP_32, LMUL_4, 8) ENTRY (RVVM2SF, TARGET_VECTOR_ELEN_FP_32, LMUL_2, 16) ENTRY (RVVM1SF, TARGET_VECTOR_ELEN_FP_32, LMUL_1, 32) -ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if !TARGET_VECTOR_ELEN_64. */ ENTRY (RVVM8DI, TARGET_VECTOR_ELEN_64, LMUL_8, 8) @@ -140,127 +140,127 @@ ENTRY (RVVM1DF, TARGET_VECTOR_ELEN_FP_64, LMUL_1, 64) #endif TUPLE_ENTRY (RVVM1x8QI, true, RVVM1QI, 8, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x8QI, true, RVVMF2QI, 8, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x8QI, true, RVVMF4QI, 8, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 8, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x8QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 8, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x8QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 8, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 8, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x7QI, true, RVVM1QI, 7, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x7QI, true, RVVMF2QI, 7, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x7QI, true, RVVMF4QI, 7, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 7, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x7QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 7, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x7QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 7, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 7, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x6QI, true, RVVM1QI, 6, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x6QI, true, RVVMF2QI, 6, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x6QI, true, RVVMF4QI, 6, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 6, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x6QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 6, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x6QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 6, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 6, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x5QI, true, RVVM1QI, 5, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x5QI, true, RVVMF2QI, 5, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x5QI, true, RVVMF4QI, 5, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 5, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x5QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 5, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x5QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 5, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 5, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x4QI, true, RVVM2QI, 4, LMUL_2, 4) TUPLE_ENTRY (RVVM1x4QI, true, RVVM1QI, 4, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x4QI, true, RVVMF2QI, 4, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x4QI, true, RVVMF4QI, 4, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 4, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x4QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 4, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x4QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 4, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 4, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x3QI, true, RVVM2QI, 3, LMUL_2, 4) TUPLE_ENTRY (RVVM1x3QI, true, RVVM1QI, 3, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x3QI, true, RVVMF2QI, 3, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x3QI, true, RVVMF4QI, 3, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 3, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x3QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 3, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x3QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 3, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 3, LMUL_F8, 64) TUPLE_ENTRY (RVVM4x2QI, true, RVVM4QI, 2, LMUL_4, 2) TUPLE_ENTRY (RVVM2x2QI, true, RVVM2QI, 2, LMUL_2, 4) TUPLE_ENTRY (RVVM1x2QI, true, RVVM1QI, 2, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x2QI, true, RVVMF2QI, 2, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x2QI, true, RVVMF4QI, 2, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 2, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x2QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 2, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x2QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 2, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 2, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x8HI, true, RVVM1HI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HI, true, RVVMF2HI, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HI, true, RVVM1HI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HI, true, RVVMF2HI, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HI, true, RVVM1HI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HI, true, RVVMF2HI, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HI, true, RVVM1HI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HI, true, RVVMF2HI, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HI, true, RVVM2HI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HI, true, RVVM1HI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HI, true, RVVMF2HI, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HI, true, RVVM2HI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HI, true, RVVM1HI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HI, true, RVVMF2HI, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HI, true, RVVM4HI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HI, true, RVVM2HI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HI, true, RVVM1HI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HI, true, RVVMF2HI, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM4HF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8SI, true, RVVM1SI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SI, true, RVVM1SI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SI, true, RVVM1SI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SI, true, RVVM1SI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SI, true, RVVM2SI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SI, true, RVVM1SI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SI, true, RVVM2SI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SI, true, RVVM1SI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SI, true, RVVM4SI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SI, true, RVVM2SI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SI, true, RVVM1SI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM4SF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 8, LMUL_1, 16) TUPLE_ENTRY (RVVM1x7DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 7, LMUL_1, 16) diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index 8f74cd2e502..28cf1f414e9 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -1406,6 +1406,9 @@ riscv_v_adjust_bytesize (machine_mode mode, int scale) { if (riscv_v_ext_vector_mode_p (mode)) { + if (TARGET_XTHEADVECTOR) + return BYTES_PER_RISCV_VECTOR; + poly_int64 nunits = GET_MODE_NUNITS (mode); poly_int64 mode_size = GET_MODE_SIZE (mode); @@ -9995,7 +9998,7 @@ riscv_use_divmod_expander (void) static machine_mode riscv_preferred_simd_mode (scalar_mode mode) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::preferred_simd_mode (mode); return word_mode; @@ -10346,7 +10349,7 @@ riscv_mode_priority (int, int n) unsigned int riscv_autovectorize_vector_modes (vector_modes *modes, bool all) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::autovectorize_vector_modes (modes, all); return default_autovectorize_vector_modes (modes, all); @@ -10529,6 +10532,16 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) return false; } +/* Implements target hook vector_mode_supported_any_target_p. */ + +static bool +riscv_vector_mode_supported_any_target_p (machine_mode mode) +{ + if (TARGET_XTHEADVECTOR) + return false; + return true; +} + /* Initialize the GCC target structure. */ #undef TARGET_ASM_ALIGNED_HI_OP #define TARGET_ASM_ALIGNED_HI_OP "\t.half\t" @@ -10872,6 +10885,9 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) #undef TARGET_PREFERRED_ELSE_VALUE #define TARGET_PREFERRED_ELSE_VALUE riscv_preferred_else_value +#undef TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P +#define TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P riscv_vector_mode_supported_any_target_p + struct gcc_target targetm = TARGET_INITIALIZER; #include "gt-riscv.h" diff --git a/gcc/config/riscv/riscv_th_vector.h b/gcc/config/riscv/riscv_th_vector.h new file mode 100644 index 00000000000..2cfcd024e76 --- /dev/null +++ b/gcc/config/riscv/riscv_th_vector.h @@ -0,0 +1,49 @@ +/* RISC-V 'XTheadVector' Extension intrinsics include file. + Copyright (C) 2023-2024 Free Software Foundation, Inc. + + This file is part of GCC. + + GCC is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published + by the Free Software Foundation; either version 3, or (at your + option) any later version. + + GCC is distributed in the hope that it will be useful, but WITHOUT + ANY WARRANTY; without even the implied warranty of MERCHANTABILITY + or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public + License for more details. + + Under Section 7 of GPL version 3, you are granted additional + permissions described in the GCC Runtime Library Exception, version + 3.1, as published by the Free Software Foundation. + + You should have received a copy of the GNU General Public License and + a copy of the GCC Runtime Library Exception along with this program; + see the files COPYING3 and COPYING.RUNTIME respectively. If not, see + <http://www.gnu.org/licenses/>. */ + +#ifndef __RISCV_TH_VECTOR_H +#define __RISCV_TH_VECTOR_H + +#include <stdint.h> +#include <stddef.h> + +#ifndef __riscv_xtheadvector +#error "XTheadVector intrinsics require the xtheadvector extension." +#else +#ifdef __cplusplus +extern "C" { +#endif + +/* NOTE: This implementation of riscv_th_vector.h is intentionally short. It does + not define the RVV types and intrinsic functions directly in C and C++ + code, but instead uses the following pragma to tell GCC to insert the + necessary type and function definitions itself. The net effect is the + same, and the file is a complete implementation of riscv_th_vector.h. */ +#pragma riscv intrinsic "xtheadvector" + +#ifdef __cplusplus +} +#endif // __cplusplus +#endif // __riscv_xtheadvector +#endif // __RISCV_TH_ECTOR_H diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md new file mode 100644 index 00000000000..696b815252d --- /dev/null +++ b/gcc/config/riscv/thead-vector.md @@ -0,0 +1,102 @@ +(define_c_enum "unspec" [ + UNSPEC_TH_VWLDST +]) + +(define_mode_iterator V_VLS_VT [V VLS VT]) +(define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) + +(define_split + [(set (match_operand:V_VB_VLS_VT 0 "reg_or_mem_operand") + (match_operand:V_VB_VLS_VT 1 "reg_or_mem_operand"))] + "TARGET_XTHEADVECTOR" + [(const_int 0)] + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + }) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:V_VLS_VT 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:V_VLS_VT + [(match_operand:V_VLS_VT 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2")]) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:VB 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:VB + [(match_operand:VB 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2") + (set (attr "sew") (const_int 8)) + (set (attr "vlmul") (symbol_ref "riscv_vector::LMUL_1"))]) + +(define_insn_and_split "*pred_th_mov<mode>" + [(set (match_operand:VB_VLS 0 "nonimmediate_operand" "=vr, m, vr, vr, vr") + (if_then_else:VB_VLS + (unspec:VB_VLS + [(match_operand:VB_VLS 1 "vector_all_trues_mask_operand" "Wc1, Wc1, Wc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) + (match_operand:VB_VLS 3 "vector_move_operand" " m, vr, vr, Wc0, Wc1") + (match_operand:VB_VLS 2 "vector_undef_operand" " vu, vu, vu, vu, vu")))] + "TARGET_XTHEADVECTOR" + "@ + # + # + vmcpy.m\t%0,%3 + vmclr.m\t%0 + vmset.m\t%0" + "&& !reload_completed" + [(const_int 0)] + { + if ((MEM_P (operands[0]) || MEM_P (operands[3])) + || (REG_P (operands[0]) && REG_P (operands[3]) + && INTVAL (operands[5]) == riscv_vector::VLMAX)) + { + emit_move_insn (operands[0], operands[3]); + DONE; + } + FAIL; + } + [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/config/riscv/thead.cc b/gcc/config/riscv/thead.cc index 00b2e14c6c0..cc04101b256 100644 --- a/gcc/config/riscv/thead.cc +++ b/gcc/config/riscv/thead.cc @@ -27,6 +27,7 @@ #include "backend.h" #include "tree.h" #include "rtl.h" +#include "insn-attr.h" #include "explow.h" #include "memmodel.h" #include "emit-rtl.h" @@ -890,8 +891,26 @@ th_asm_output_opcode (FILE *asm_out_file, const char *p) { /* We need to add th. prefix to all the xtheadvector instructions here.*/ - if (current_output_insn != NULL && p[0] == 'v') - fputs ("th.", asm_out_file); + if (current_output_insn != NULL) + { + if (get_attr_type (current_output_insn) == TYPE_VSETVL) + { + if (strstr (p, "zero")) + { + if (strstr (p, "zero,zero")) + return "th.vsetvli\tzero,zero,e%0,%m1"; + else + return "th.vsetvli\tzero,%0,e%1,%m2"; + } + else + { + return "th.vsetvli\t%0,%1,e%2,%m3"; + } + } + + if (p[0] == 'v') + fputs ("th.", asm_out_file); + } return p; } diff --git a/gcc/config/riscv/vector-iterators.md b/gcc/config/riscv/vector-iterators.md index b4a276dc2c8..a993487fd88 100644 --- a/gcc/config/riscv/vector-iterators.md +++ b/gcc/config/riscv/vector-iterators.md @@ -109,11 +109,11 @@ ]) (define_mode_iterator VI [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -128,11 +128,11 @@ ;; allow the instruction and mode to be matched during combine et al. (define_mode_iterator VF [ (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -140,11 +140,11 @@ (define_mode_iterator VF_ZVFHMIN [ (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -271,16 +271,16 @@ ]) (define_mode_iterator VEEWEXT2 [ - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -290,10 +290,10 @@ ]) (define_mode_iterator VEEWEXT4 [ - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -311,59 +311,59 @@ ]) (define_mode_iterator VEEWTRUNC2 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") (RVVM4SI "TARGET_64BIT") (RVVM2SI "TARGET_64BIT") (RVVM1SI "TARGET_64BIT") - (RVVMF2SI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEEWTRUNC4 [ - RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM2HI "TARGET_64BIT") (RVVM1HI "TARGET_64BIT") - (RVVMF2HI "TARGET_64BIT") - (RVVMF4HI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2HI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") (RVVM1HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEEWTRUNC8 [ (RVVM1QI "TARGET_64BIT") - (RVVMF2QI "TARGET_64BIT") - (RVVMF4QI "TARGET_64BIT") - (RVVMF8QI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2QI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF4QI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEI16 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -452,11 +452,11 @@ ]) (define_mode_iterator VFULLI [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_FULL_V") (RVVM4DI "TARGET_FULL_V") (RVVM2DI "TARGET_FULL_V") (RVVM1DI "TARGET_FULL_V") @@ -509,17 +509,17 @@ ]) (define_mode_iterator VI_QH [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") ]) (define_mode_iterator VI_QHS [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -560,11 +560,11 @@ ]) (define_mode_iterator VI_QHS_NO_M8 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -603,11 +603,11 @@ (define_mode_iterator VF_HS [ (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (V1HF "riscv_vector::vls_mode_valid_p (V1HFmode) && TARGET_ZVFH") (V2HF "riscv_vector::vls_mode_valid_p (V2HFmode) && TARGET_ZVFH") @@ -638,12 +638,12 @@ (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") - (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (V1HF "riscv_vector::vls_mode_valid_p (V1HFmode) && TARGET_ZVFH") (V2HF "riscv_vector::vls_mode_valid_p (V2HFmode) && TARGET_ZVFH") @@ -674,11 +674,11 @@ ]) (define_mode_iterator V_VLSI_QHS [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -756,27 +756,27 @@ ;; E.g. when index mode = RVVM8QImde and Pmode = SImode, if it is not zero_extend or ;; scalar != 1, such gather/scatter is not allowed since we don't have RVVM32SImode. (define_mode_iterator RATIO64 [ - (RVVMF8QI "TARGET_MIN_VLEN > 32") - (RVVMF4HI "TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM1DI "TARGET_VECTOR_ELEN_64") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") ]) (define_mode_iterator RATIO32 [ - RVVMF4QI - RVVMF2HI + (RVVMF4QI "!TARGET_XTHEADVECTOR") + (RVVMF2HI "!TARGET_XTHEADVECTOR") RVVM1SI (RVVM2DI "TARGET_VECTOR_ELEN_64") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") ]) (define_mode_iterator RATIO16 [ - RVVMF2QI + (RVVMF2QI "!TARGET_XTHEADVECTOR") RVVM1HI RVVM2SI (RVVM4DI "TARGET_VECTOR_ELEN_64") @@ -814,21 +814,21 @@ ]) (define_mode_iterator RATIO64I [ - (RVVMF8QI "TARGET_MIN_VLEN > 32") - (RVVMF4HI "TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM1DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") ]) (define_mode_iterator RATIO32I [ - RVVMF4QI - RVVMF2HI + (RVVMF4QI "!TARGET_XTHEADVECTOR") + (RVVMF2HI "!TARGET_XTHEADVECTOR") RVVM1SI (RVVM2DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") ]) (define_mode_iterator RATIO16I [ - RVVMF2QI + (RVVMF2QI "!TARGET_XTHEADVECTOR") RVVM1HI RVVM2SI (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") @@ -873,21 +873,21 @@ ]) (define_mode_iterator V_FRACT [ - RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") ]) (define_mode_iterator VWEXTI [ - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -933,7 +933,7 @@ (RVVM4SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -966,7 +966,7 @@ (RVVM4SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -996,7 +996,7 @@ (define_mode_iterator VWCONVERTI [ (RVVM8SI "TARGET_ZVFH") (RVVM4SI "TARGET_ZVFH") (RVVM2SI "TARGET_ZVFH") (RVVM1SI "TARGET_ZVFH") - (RVVMF2SI "TARGET_ZVFH") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_ZVFH") (RVVM8DI "TARGET_VECTOR_ELEN_64 && TARGET_VECTOR_ELEN_FP_32") (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_VECTOR_ELEN_FP_32") @@ -1045,7 +1045,7 @@ ]) (define_mode_iterator VQEXTI [ - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -1456,11 +1456,11 @@ ;; VINDEXED [VI8 VI16 VI32 (VI64 "TARGET_64BIT")]. (define_mode_iterator VINDEXED [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") @@ -1468,12 +1468,12 @@ (RVVM1DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64 && TARGET_64BIT") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64 && TARGET_64BIT") @@ -3173,11 +3173,11 @@ (define_mode_iterator V_VLS_F_CONVERT_SI [ (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") - (RVVMF2HF "TARGET_ZVFH") (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") @@ -3290,12 +3290,12 @@ ]) (define_mode_iterator V_VLS_F_CONVERT_DI [ - (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index 24b7b4394be..3eb6daafbc2 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -87,7 +87,7 @@ ;; check. However, we need default value of SEW for vsetvl instruction since there ;; is no field for ratio in the vsetvl instruction encoding. (define_attr "sew" "" - (cond [(eq_attr "mode" "RVVMF64BI,RVVMF32BI,RVVMF16BI,RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ + (cond [(eq_attr "mode" "RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ RVVM8QI,RVVM4QI,RVVM2QI,RVVM1QI,RVVMF2QI,RVVMF4QI,RVVMF8QI,\ RVVM1x8QI,RVVMF2x8QI,RVVMF4x8QI,RVVMF8x8QI,\ RVVM1x7QI,RVVMF2x7QI,RVVMF4x7QI,RVVMF8x7QI,\ @@ -99,6 +99,18 @@ V1QI,V2QI,V4QI,V8QI,V16QI,V32QI,V64QI,V128QI,V256QI,V512QI,V1024QI,V2048QI,V4096QI,\ V1BI,V2BI,V4BI,V8BI,V16BI,V32BI,V64BI,V128BI,V256BI,V512BI,V1024BI,V2048BI,V4096BI") (const_int 8) + (eq_attr "mode" "RVVMF16BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 16) + (const_int 8)) + (eq_attr "mode" "RVVMF32BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 32) + (const_int 8)) + (eq_attr "mode" "RVVMF64BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 64) + (const_int 8)) (eq_attr "mode" "RVVM8HI,RVVM4HI,RVVM2HI,RVVM1HI,RVVMF2HI,RVVMF4HI,\ RVVM1x8HI,RVVMF2x8HI,RVVMF4x8HI,\ RVVM1x7HI,RVVMF2x7HI,RVVMF4x7HI,\ @@ -159,9 +171,9 @@ (eq_attr "mode" "RVVM4QI,RVVMF2BI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2QI,RVVMF4BI") (symbol_ref "riscv_vector::LMUL_2") (eq_attr "mode" "RVVM1QI,RVVMF8BI") (symbol_ref "riscv_vector::LMUL_1") - (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "riscv_vector::LMUL_F2") - (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "riscv_vector::LMUL_F4") - (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "riscv_vector::LMUL_F8") + (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F2") + (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F4") + (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F8") (eq_attr "mode" "RVVM8HI") (symbol_ref "riscv_vector::LMUL_8") (eq_attr "mode" "RVVM4HI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2HI") (symbol_ref "riscv_vector::LMUL_2") @@ -436,6 +448,10 @@ vaeskf1,vaeskf2,vaesz,vsha2ms,vsha2ch,vsha2cl,vsm4k,vsm4r,\ vsm3me,vsm3c") (const_int INVALID_ATTRIBUTE) + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE) (eq_attr "mode" "RVVM8QI,RVVM1BI") (const_int 1) (eq_attr "mode" "RVVM4QI,RVVMF2BI") (const_int 2) (eq_attr "mode" "RVVM2QI,RVVMF4BI") (const_int 4) @@ -905,6 +921,8 @@ (symbol_ref "riscv_vector::FRM_DYN")] (symbol_ref "riscv_vector::FRM_NONE"))) +(include "thead-vector.md") + ;; ----------------------------------------------------------------- ;; ---- Miscellaneous Operations ;; ----------------------------------------------------------------- @@ -1078,6 +1096,12 @@ - We can not leave it to TARGET_SECONDARY_RELOAD since it happens before spilling. The clobber scratch is used by spilling fractional registers in IRA/LRA so it's too early. */ + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; @@ -1114,7 +1138,7 @@ (define_insn "*mov<mode>_whole" [(set (match_operand:V_WHOLE 0 "reg_or_mem_operand" "=vr, m,vr") (match_operand:V_WHOLE 1 "reg_or_mem_operand" " m,vr,vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "@ vl%m1re<sew>.v\t%0,%1 vs%m1r.v\t%1,%0 @@ -1135,6 +1159,13 @@ (match_operand:VB 1 "general_operand"))] "TARGET_VECTOR" { + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } + if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; }) @@ -1142,7 +1173,7 @@ (define_insn "*mov<mode>" [(set (match_operand:VB 0 "register_operand" "=vr") (match_operand:VB 1 "register_operand" " vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "vmv1r.v\t%0,%1" [(set_attr "type" "vmov") (set_attr "mode" "<MODE>")]) @@ -3692,7 +3723,7 @@ (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3713,7 +3744,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3734,7 +3765,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c index 2e0e12aa045..2eef9e1e1a8 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c @@ -1,4 +1,4 @@ -/* { dg-do compile } */ +/* { dg-do compile { target { ! riscv_xtheadvector } } } */ /* { dg-skip-if "test rvv intrinsic" { *-*-* } { "*" } { "-march=rv*v*" } } */ void foo0 () {__rvv_bool64_t t;} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c index 3d81b179235..ef329e30785 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c @@ -1,4 +1,4 @@ /* { dg-do compile } */ /* { dg-options "-O3 -march=rv32gc -mabi=ilp32d" } */ -#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' extension enabled} } */ +#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' or 'XTHEADVECTOR' extension enabled} } */ diff --git a/gcc/testsuite/lib/target-supports.exp b/gcc/testsuite/lib/target-supports.exp index b27c30b8c51..bf0ad5342d6 100644 --- a/gcc/testsuite/lib/target-supports.exp +++ b/gcc/testsuite/lib/target-supports.exp @@ -1956,6 +1956,18 @@ proc check_effective_target_riscv_zbb { } { }] } +# Return 1 if the target arch supports the XTheadVector extension, 0 otherwise. +# Cache the result. + +proc check_effective_target_riscv_xtheadvector { } { + return [check_no_compiler_messages riscv_ext_xtheadvector assembly { + #ifndef __riscv_xtheadvector + #error "Not __riscv_xtheadvector" + #endif + }] +} + + # Return 1 if we can execute code when using dg-add-options riscv_v proc check_effective_target_riscv_v_ok { } { -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: [PATCH v5] RISC-V: Handle differences between XTheadvector and Vector 2024-01-09 3:18 ` [PATCH v5] " Jun Sha (Joshua) @ 2024-01-09 22:33 ` 钟居哲 2024-01-10 2:22 ` Jun Sha (Joshua) 1 sibling, 0 replies; 133+ messages in thread From: 钟居哲 @ 2024-01-09 22:33 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, Jeff Law, Christoph Müllner, cooper.joshua, jinma, Cooper Qu [-- Attachment #1: Type: text/plain, Size: 70290 bytes --] + Copyright (C) 2023-2024 Free Software Foundation, Inc. It should be: Copyright (C) 2024 Free Software Foundation, Inc. (define_mode_iterator VI [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") You have blocked tuple modes or fractional modes in riscv-vector-switch.def, that is blocking those modes in riscv_vector_mode_supported_p targethook. I think we don't need to modify vector-iterators.md juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-09 11:18 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Handle differences between XTheadvector and Vector This patch is to handle the differences in instruction generation between Vector and XTheadVector. In this version, we only support partial xtheadvector instructions that leverage directly from current RVV1.0 with simple adding "th." prefix. For different name xtheadvector instructions but share same patterns as RVV1.0 instructions, we will use ASM targethook to rewrite the whole string of the instructions in the following patches. For some vector patterns that cannot be avoided, we use "!TARGET_XTHEADVECTOR" to disable them in vector.md in order not to generate instructions that xtheadvector does not support, like vmv1r and vsext.vf2. gcc/ChangeLog: * config.gcc: Add files for XTheadVector intrinsics. * config/riscv/autovec.md: Guard XTheadVector. * config/riscv/riscv-c.cc: Add pragma for XTheadVector. * config/riscv/riscv-string.cc (expand_block_move): Guard XTheadVector. (get_prefer_tail_policy): Give specific value for tail. (get_prefer_mask_policy): Give specific value for mask. (vls_mode_valid_p): Avoid autovec. * config/riscv/riscv-vector-builtins-shapes.cc (check_type): (build_one): New function. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): (DEF_THEAD_RVV_FUNCTION): Add new marcos. (check_required_extensions): (handle_pragma_vector): * config/riscv/riscv-vector-builtins.h (RVV_REQUIRE_VECTOR): (RVV_REQUIRE_XTHEADVECTOR): Add RVV_REQUIRE_VECTOR and RVV_REQUIRE_XTHEADVECTOR. (struct function_group_info): * config/riscv/riscv-vector-switch.def (ENTRY): Disable fractional mode for the XTheadVector extension. (TUPLE_ENTRY): Likewise. * config/riscv/riscv-vsetvl.cc: Add functions for xtheadvector. * config/riscv/riscv.cc (riscv_v_ext_vls_mode_p): Guard XTheadVector. (riscv_v_adjust_bytesize): Likewise. (riscv_preferred_simd_mode): Likewsie. (riscv_autovectorize_vector_modes): Likewise. (riscv_vector_mode_supported_any_target_p): Likewise. (TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P): Likewise. * config/riscv/vector-iterators.md: Remove fractional LMUL. * config/riscv/vector.md: Include thead-vector.md. * config/riscv/riscv_th_vector.h: New file. * config/riscv/thead-vector.md: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/base/pragma-1.c: Add XTheadVector. * gcc.target/riscv/rvv/base/abi-1.c: Exclude XTheadVector. * lib/target-supports.exp: Add target for XTheadVector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config.gcc | 2 +- gcc/config/riscv/autovec.md | 2 +- gcc/config/riscv/predicates.md | 4 +- gcc/config/riscv/riscv-c.cc | 3 +- gcc/config/riscv/riscv-string.cc | 3 +- gcc/config/riscv/riscv-v.cc | 2 +- .../riscv/riscv-vector-builtins-bases.cc | 48 +++-- .../riscv/riscv-vector-builtins-shapes.cc | 23 +++ gcc/config/riscv/riscv-vector-switch.def | 150 +++++++------- gcc/config/riscv/riscv.cc | 20 +- gcc/config/riscv/riscv_th_vector.h | 49 +++++ gcc/config/riscv/thead-vector.md | 102 ++++++++++ gcc/config/riscv/thead.cc | 23 ++- gcc/config/riscv/vector-iterators.md | 186 +++++++++--------- gcc/config/riscv/vector.md | 49 ++++- .../gcc.target/riscv/rvv/base/abi-1.c | 2 +- .../gcc.target/riscv/rvv/base/pragma-1.c | 2 +- gcc/testsuite/lib/target-supports.exp | 12 ++ 18 files changed, 476 insertions(+), 206 deletions(-) create mode 100644 gcc/config/riscv/riscv_th_vector.h create mode 100644 gcc/config/riscv/thead-vector.md diff --git a/gcc/config.gcc b/gcc/config.gcc index 7e583390024..047e4c02cf4 100644 --- a/gcc/config.gcc +++ b/gcc/config.gcc @@ -549,7 +549,7 @@ riscv*) extra_objs="${extra_objs} riscv-vector-builtins.o riscv-vector-builtins-shapes.o riscv-vector-builtins-bases.o" extra_objs="${extra_objs} thead.o riscv-target-attr.o" d_target_objs="riscv-d.o" - extra_headers="riscv_vector.h" + extra_headers="riscv_vector.h riscv_th_vector.h" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.cc" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.h" ;; diff --git a/gcc/config/riscv/autovec.md b/gcc/config/riscv/autovec.md index 775eaa825b0..0477781cabe 100644 --- a/gcc/config/riscv/autovec.md +++ b/gcc/config/riscv/autovec.md @@ -2579,7 +2579,7 @@ [(match_operand 0 "register_operand") (match_operand 1 "memory_operand") (match_operand:ANYI 2 "const_int_operand")] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" { riscv_vector::expand_rawmemchr(<MODE>mode, operands[0], operands[1], operands[2]); diff --git a/gcc/config/riscv/predicates.md b/gcc/config/riscv/predicates.md index b1a79cae50a..0337da88284 100644 --- a/gcc/config/riscv/predicates.md +++ b/gcc/config/riscv/predicates.md @@ -428,7 +428,9 @@ ;; Predicates for the V extension. (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (ior (match_test "TARGET_XTHEADVECTOR && rtx_equal_p (op, const0_rtx)") + (match_test "!TARGET_XTHEADVECTOR")) + (match_operand 0 "const_csr_operand")))) (define_special_predicate "autovec_length_operand" (ior (match_operand 0 "pmode_register_operand") diff --git a/gcc/config/riscv/riscv-c.cc b/gcc/config/riscv/riscv-c.cc index 422ddc2c308..2e306057347 100644 --- a/gcc/config/riscv/riscv-c.cc +++ b/gcc/config/riscv/riscv-c.cc @@ -195,7 +195,8 @@ riscv_pragma_intrinsic (cpp_reader *) const char *name = TREE_STRING_POINTER (x); - if (strcmp (name, "vector") == 0) + if (strcmp (name, "vector") == 0 + || strcmp (name, "xtheadvector") == 0) { if (!TARGET_VECTOR) { diff --git a/gcc/config/riscv/riscv-string.cc b/gcc/config/riscv/riscv-string.cc index f227b8667ce..b09b51d7526 100644 --- a/gcc/config/riscv/riscv-string.cc +++ b/gcc/config/riscv/riscv-string.cc @@ -773,7 +773,8 @@ riscv_expand_block_move_scalar (rtx dest, rtx src, rtx length) bool riscv_expand_block_move (rtx dest, rtx src, rtx length) { - if (TARGET_VECTOR && stringop_strategy & STRATEGY_VECTOR) + if ((TARGET_VECTOR && !TARGET_XTHEADVECTOR) + && stringop_strategy & STRATEGY_VECTOR) { bool ok = riscv_vector::expand_block_move (dest, src, length); if (ok) diff --git a/gcc/config/riscv/riscv-v.cc b/gcc/config/riscv/riscv-v.cc index 2491522191a..7e25deb966d 100644 --- a/gcc/config/riscv/riscv-v.cc +++ b/gcc/config/riscv/riscv-v.cc @@ -4392,7 +4392,7 @@ cmp_lmul_gt_one (machine_mode mode) bool vls_mode_valid_p (machine_mode vls_mode) { - if (!TARGET_VECTOR) + if (!TARGET_VECTOR || TARGET_XTHEADVECTOR) return false; if (riscv_autovec_preference == RVV_SCALABLE) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 91af2c00674..46f1a1da33e 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -115,23 +115,37 @@ public: tree type = builtin_types[e.type.index].vector; machine_mode mode = TYPE_MODE (type); - /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. - - - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 - - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 - - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 - - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 - - e8,m2/e16,m4/e32,m8 --> e8m2 - - e8,m4/e16,m8 --> e8m4 - - e8,m8 --> e8m8 - */ - /* SEW. */ - e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); - - /* LMUL. */ - machine_mode e8_mode - = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); - e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + + if (TARGET_XTHEADVECTOR) + { + machine_mode inner_mode = GET_MODE_INNER (mode); + /* SEW. */ + e.add_input_operand (Pmode, + gen_int_mode (GET_MODE_BITSIZE (inner_mode), Pmode)); + /* LMUL. */ + e.add_input_operand (Pmode, + gen_int_mode (get_vlmul (mode), Pmode)); + } + else + { + /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. + + - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 + - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 + - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 + - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 + - e8,m2/e16,m4/e32,m8 --> e8m2 + - e8,m4/e16,m8 --> e8m4 + - e8,m8 --> e8m8 + */ + /* SEW. */ + e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); + + /* LMUL. */ + machine_mode e8_mode + = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); + e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + } /* TAIL_ANY. */ e.add_input_operand (Pmode, diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index ee8058dcdb1..1e4f4d53de6 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -33,6 +33,25 @@ namespace riscv_vector { +/* Check whether the RETURN_TYPE and ARGUMENT_TYPES are + valid for the function. */ + +static bool +check_type (tree return_type, vec<tree> &argument_types) +{ + tree arg; + unsigned i; + + if (!return_type) + return false; + + FOR_EACH_VEC_ELT (argument_types, i, arg) + if (!arg) + return false; + + return true; +} + /* Add one function instance for GROUP, using operand suffix at index OI, mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ static void @@ -49,6 +68,10 @@ build_one (function_builder &b, const function_group_info &group, group.ops_infos.types[vec_type_idx].index); b.allocate_argument_types (function_instance, argument_types); b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + b.add_overloaded_function (function_instance, *group.shape); b.add_unique_function (function_instance, (*group.shape), return_type, argument_types); diff --git a/gcc/config/riscv/riscv-vector-switch.def b/gcc/config/riscv/riscv-vector-switch.def index 1ad26c2a3b2..452283b7416 100644 --- a/gcc/config/riscv/riscv-vector-switch.def +++ b/gcc/config/riscv/riscv-vector-switch.def @@ -68,9 +68,9 @@ Encode the ratio of SEW/LMUL into the mask types. #endif /* Disable modes if TARGET_MIN_VLEN == 32. */ -ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) -ENTRY (RVVMF32BI, true, LMUL_F4, 32) -ENTRY (RVVMF16BI, true, LMUL_F2, 16) +ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F8, 64) +ENTRY (RVVMF32BI, true, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F4, 32) +ENTRY (RVVMF16BI, true, TARGET_XTHEADVECTOR ? LMUL_1 : LMUL_F2 , 16) ENTRY (RVVMF8BI, true, LMUL_1, 8) ENTRY (RVVMF4BI, true, LMUL_2, 4) ENTRY (RVVMF2BI, true, LMUL_4, 2) @@ -81,39 +81,39 @@ ENTRY (RVVM8QI, true, LMUL_8, 1) ENTRY (RVVM4QI, true, LMUL_4, 2) ENTRY (RVVM2QI, true, LMUL_2, 4) ENTRY (RVVM1QI, true, LMUL_1, 8) -ENTRY (RVVMF2QI, true, LMUL_F2, 16) -ENTRY (RVVMF4QI, true, LMUL_F4, 32) -ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) +ENTRY (RVVMF2QI, !TARGET_XTHEADVECTOR, LMUL_F2, 16) +ENTRY (RVVMF4QI, !TARGET_XTHEADVECTOR, LMUL_F4, 32) +ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F8, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8HI, true, LMUL_8, 2) ENTRY (RVVM4HI, true, LMUL_4, 4) ENTRY (RVVM2HI, true, LMUL_2, 8) ENTRY (RVVM1HI, true, LMUL_1, 16) -ENTRY (RVVMF2HI, true, LMUL_F2, 32) -ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HI, !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_16. */ ENTRY (RVVM8HF, TARGET_VECTOR_ELEN_FP_16, LMUL_8, 2) ENTRY (RVVM4HF, TARGET_VECTOR_ELEN_FP_16, LMUL_4, 4) ENTRY (RVVM2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_2, 8) ENTRY (RVVM1HF, TARGET_VECTOR_ELEN_FP_16, LMUL_1, 16) -ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_F2, 32) -ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8SI, true, LMUL_8, 4) ENTRY (RVVM4SI, true, LMUL_4, 8) ENTRY (RVVM2SI, true, LMUL_2, 16) ENTRY (RVVM1SI, true, LMUL_1, 32) -ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_32. */ ENTRY (RVVM8SF, TARGET_VECTOR_ELEN_FP_32, LMUL_8, 4) ENTRY (RVVM4SF, TARGET_VECTOR_ELEN_FP_32, LMUL_4, 8) ENTRY (RVVM2SF, TARGET_VECTOR_ELEN_FP_32, LMUL_2, 16) ENTRY (RVVM1SF, TARGET_VECTOR_ELEN_FP_32, LMUL_1, 32) -ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if !TARGET_VECTOR_ELEN_64. */ ENTRY (RVVM8DI, TARGET_VECTOR_ELEN_64, LMUL_8, 8) @@ -140,127 +140,127 @@ ENTRY (RVVM1DF, TARGET_VECTOR_ELEN_FP_64, LMUL_1, 64) #endif TUPLE_ENTRY (RVVM1x8QI, true, RVVM1QI, 8, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x8QI, true, RVVMF2QI, 8, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x8QI, true, RVVMF4QI, 8, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 8, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x8QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 8, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x8QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 8, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 8, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x7QI, true, RVVM1QI, 7, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x7QI, true, RVVMF2QI, 7, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x7QI, true, RVVMF4QI, 7, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 7, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x7QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 7, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x7QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 7, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 7, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x6QI, true, RVVM1QI, 6, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x6QI, true, RVVMF2QI, 6, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x6QI, true, RVVMF4QI, 6, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 6, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x6QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 6, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x6QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 6, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 6, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x5QI, true, RVVM1QI, 5, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x5QI, true, RVVMF2QI, 5, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x5QI, true, RVVMF4QI, 5, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 5, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x5QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 5, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x5QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 5, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 5, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x4QI, true, RVVM2QI, 4, LMUL_2, 4) TUPLE_ENTRY (RVVM1x4QI, true, RVVM1QI, 4, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x4QI, true, RVVMF2QI, 4, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x4QI, true, RVVMF4QI, 4, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 4, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x4QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 4, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x4QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 4, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 4, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x3QI, true, RVVM2QI, 3, LMUL_2, 4) TUPLE_ENTRY (RVVM1x3QI, true, RVVM1QI, 3, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x3QI, true, RVVMF2QI, 3, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x3QI, true, RVVMF4QI, 3, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 3, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x3QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 3, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x3QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 3, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 3, LMUL_F8, 64) TUPLE_ENTRY (RVVM4x2QI, true, RVVM4QI, 2, LMUL_4, 2) TUPLE_ENTRY (RVVM2x2QI, true, RVVM2QI, 2, LMUL_2, 4) TUPLE_ENTRY (RVVM1x2QI, true, RVVM1QI, 2, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x2QI, true, RVVMF2QI, 2, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x2QI, true, RVVMF4QI, 2, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 2, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x2QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 2, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x2QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 2, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 2, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x8HI, true, RVVM1HI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HI, true, RVVMF2HI, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HI, true, RVVM1HI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HI, true, RVVMF2HI, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HI, true, RVVM1HI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HI, true, RVVMF2HI, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HI, true, RVVM1HI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HI, true, RVVMF2HI, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HI, true, RVVM2HI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HI, true, RVVM1HI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HI, true, RVVMF2HI, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HI, true, RVVM2HI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HI, true, RVVM1HI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HI, true, RVVMF2HI, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HI, true, RVVM4HI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HI, true, RVVM2HI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HI, true, RVVM1HI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HI, true, RVVMF2HI, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM4HF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8SI, true, RVVM1SI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SI, true, RVVM1SI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SI, true, RVVM1SI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SI, true, RVVM1SI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SI, true, RVVM2SI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SI, true, RVVM1SI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SI, true, RVVM2SI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SI, true, RVVM1SI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SI, true, RVVM4SI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SI, true, RVVM2SI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SI, true, RVVM1SI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM4SF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 8, LMUL_1, 16) TUPLE_ENTRY (RVVM1x7DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 7, LMUL_1, 16) diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index 8f74cd2e502..28cf1f414e9 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -1406,6 +1406,9 @@ riscv_v_adjust_bytesize (machine_mode mode, int scale) { if (riscv_v_ext_vector_mode_p (mode)) { + if (TARGET_XTHEADVECTOR) + return BYTES_PER_RISCV_VECTOR; + poly_int64 nunits = GET_MODE_NUNITS (mode); poly_int64 mode_size = GET_MODE_SIZE (mode); @@ -9995,7 +9998,7 @@ riscv_use_divmod_expander (void) static machine_mode riscv_preferred_simd_mode (scalar_mode mode) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::preferred_simd_mode (mode); return word_mode; @@ -10346,7 +10349,7 @@ riscv_mode_priority (int, int n) unsigned int riscv_autovectorize_vector_modes (vector_modes *modes, bool all) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::autovectorize_vector_modes (modes, all); return default_autovectorize_vector_modes (modes, all); @@ -10529,6 +10532,16 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) return false; } +/* Implements target hook vector_mode_supported_any_target_p. */ + +static bool +riscv_vector_mode_supported_any_target_p (machine_mode mode) +{ + if (TARGET_XTHEADVECTOR) + return false; + return true; +} + /* Initialize the GCC target structure. */ #undef TARGET_ASM_ALIGNED_HI_OP #define TARGET_ASM_ALIGNED_HI_OP "\t.half\t" @@ -10872,6 +10885,9 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) #undef TARGET_PREFERRED_ELSE_VALUE #define TARGET_PREFERRED_ELSE_VALUE riscv_preferred_else_value +#undef TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P +#define TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P riscv_vector_mode_supported_any_target_p + struct gcc_target targetm = TARGET_INITIALIZER; #include "gt-riscv.h" diff --git a/gcc/config/riscv/riscv_th_vector.h b/gcc/config/riscv/riscv_th_vector.h new file mode 100644 index 00000000000..2cfcd024e76 --- /dev/null +++ b/gcc/config/riscv/riscv_th_vector.h @@ -0,0 +1,49 @@ +/* RISC-V 'XTheadVector' Extension intrinsics include file. + Copyright (C) 2023-2024 Free Software Foundation, Inc. + + This file is part of GCC. + + GCC is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published + by the Free Software Foundation; either version 3, or (at your + option) any later version. + + GCC is distributed in the hope that it will be useful, but WITHOUT + ANY WARRANTY; without even the implied warranty of MERCHANTABILITY + or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public + License for more details. + + Under Section 7 of GPL version 3, you are granted additional + permissions described in the GCC Runtime Library Exception, version + 3.1, as published by the Free Software Foundation. + + You should have received a copy of the GNU General Public License and + a copy of the GCC Runtime Library Exception along with this program; + see the files COPYING3 and COPYING.RUNTIME respectively. If not, see + <http://www.gnu.org/licenses/>. */ + +#ifndef __RISCV_TH_VECTOR_H +#define __RISCV_TH_VECTOR_H + +#include <stdint.h> +#include <stddef.h> + +#ifndef __riscv_xtheadvector +#error "XTheadVector intrinsics require the xtheadvector extension." +#else +#ifdef __cplusplus +extern "C" { +#endif + +/* NOTE: This implementation of riscv_th_vector.h is intentionally short. It does + not define the RVV types and intrinsic functions directly in C and C++ + code, but instead uses the following pragma to tell GCC to insert the + necessary type and function definitions itself. The net effect is the + same, and the file is a complete implementation of riscv_th_vector.h. */ +#pragma riscv intrinsic "xtheadvector" + +#ifdef __cplusplus +} +#endif // __cplusplus +#endif // __riscv_xtheadvector +#endif // __RISCV_TH_ECTOR_H diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md new file mode 100644 index 00000000000..696b815252d --- /dev/null +++ b/gcc/config/riscv/thead-vector.md @@ -0,0 +1,102 @@ +(define_c_enum "unspec" [ + UNSPEC_TH_VWLDST +]) + +(define_mode_iterator V_VLS_VT [V VLS VT]) +(define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) + +(define_split + [(set (match_operand:V_VB_VLS_VT 0 "reg_or_mem_operand") + (match_operand:V_VB_VLS_VT 1 "reg_or_mem_operand"))] + "TARGET_XTHEADVECTOR" + [(const_int 0)] + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + }) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:V_VLS_VT 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:V_VLS_VT + [(match_operand:V_VLS_VT 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2")]) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:VB 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:VB + [(match_operand:VB 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2") + (set (attr "sew") (const_int 8)) + (set (attr "vlmul") (symbol_ref "riscv_vector::LMUL_1"))]) + +(define_insn_and_split "*pred_th_mov<mode>" + [(set (match_operand:VB_VLS 0 "nonimmediate_operand" "=vr, m, vr, vr, vr") + (if_then_else:VB_VLS + (unspec:VB_VLS + [(match_operand:VB_VLS 1 "vector_all_trues_mask_operand" "Wc1, Wc1, Wc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) + (match_operand:VB_VLS 3 "vector_move_operand" " m, vr, vr, Wc0, Wc1") + (match_operand:VB_VLS 2 "vector_undef_operand" " vu, vu, vu, vu, vu")))] + "TARGET_XTHEADVECTOR" + "@ + # + # + vmcpy.m\t%0,%3 + vmclr.m\t%0 + vmset.m\t%0" + "&& !reload_completed" + [(const_int 0)] + { + if ((MEM_P (operands[0]) || MEM_P (operands[3])) + || (REG_P (operands[0]) && REG_P (operands[3]) + && INTVAL (operands[5]) == riscv_vector::VLMAX)) + { + emit_move_insn (operands[0], operands[3]); + DONE; + } + FAIL; + } + [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/config/riscv/thead.cc b/gcc/config/riscv/thead.cc index 00b2e14c6c0..cc04101b256 100644 --- a/gcc/config/riscv/thead.cc +++ b/gcc/config/riscv/thead.cc @@ -27,6 +27,7 @@ #include "backend.h" #include "tree.h" #include "rtl.h" +#include "insn-attr.h" #include "explow.h" #include "memmodel.h" #include "emit-rtl.h" @@ -890,8 +891,26 @@ th_asm_output_opcode (FILE *asm_out_file, const char *p) { /* We need to add th. prefix to all the xtheadvector instructions here.*/ - if (current_output_insn != NULL && p[0] == 'v') - fputs ("th.", asm_out_file); + if (current_output_insn != NULL) + { + if (get_attr_type (current_output_insn) == TYPE_VSETVL) + { + if (strstr (p, "zero")) + { + if (strstr (p, "zero,zero")) + return "th.vsetvli\tzero,zero,e%0,%m1"; + else + return "th.vsetvli\tzero,%0,e%1,%m2"; + } + else + { + return "th.vsetvli\t%0,%1,e%2,%m3"; + } + } + + if (p[0] == 'v') + fputs ("th.", asm_out_file); + } return p; } diff --git a/gcc/config/riscv/vector-iterators.md b/gcc/config/riscv/vector-iterators.md index b4a276dc2c8..a993487fd88 100644 --- a/gcc/config/riscv/vector-iterators.md +++ b/gcc/config/riscv/vector-iterators.md @@ -109,11 +109,11 @@ ]) (define_mode_iterator VI [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -128,11 +128,11 @@ ;; allow the instruction and mode to be matched during combine et al. (define_mode_iterator VF [ (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -140,11 +140,11 @@ (define_mode_iterator VF_ZVFHMIN [ (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -271,16 +271,16 @@ ]) (define_mode_iterator VEEWEXT2 [ - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -290,10 +290,10 @@ ]) (define_mode_iterator VEEWEXT4 [ - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -311,59 +311,59 @@ ]) (define_mode_iterator VEEWTRUNC2 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") (RVVM4SI "TARGET_64BIT") (RVVM2SI "TARGET_64BIT") (RVVM1SI "TARGET_64BIT") - (RVVMF2SI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEEWTRUNC4 [ - RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM2HI "TARGET_64BIT") (RVVM1HI "TARGET_64BIT") - (RVVMF2HI "TARGET_64BIT") - (RVVMF4HI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2HI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") (RVVM1HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEEWTRUNC8 [ (RVVM1QI "TARGET_64BIT") - (RVVMF2QI "TARGET_64BIT") - (RVVMF4QI "TARGET_64BIT") - (RVVMF8QI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2QI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF4QI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEI16 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -452,11 +452,11 @@ ]) (define_mode_iterator VFULLI [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_FULL_V") (RVVM4DI "TARGET_FULL_V") (RVVM2DI "TARGET_FULL_V") (RVVM1DI "TARGET_FULL_V") @@ -509,17 +509,17 @@ ]) (define_mode_iterator VI_QH [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") ]) (define_mode_iterator VI_QHS [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -560,11 +560,11 @@ ]) (define_mode_iterator VI_QHS_NO_M8 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -603,11 +603,11 @@ (define_mode_iterator VF_HS [ (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (V1HF "riscv_vector::vls_mode_valid_p (V1HFmode) && TARGET_ZVFH") (V2HF "riscv_vector::vls_mode_valid_p (V2HFmode) && TARGET_ZVFH") @@ -638,12 +638,12 @@ (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") - (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (V1HF "riscv_vector::vls_mode_valid_p (V1HFmode) && TARGET_ZVFH") (V2HF "riscv_vector::vls_mode_valid_p (V2HFmode) && TARGET_ZVFH") @@ -674,11 +674,11 @@ ]) (define_mode_iterator V_VLSI_QHS [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -756,27 +756,27 @@ ;; E.g. when index mode = RVVM8QImde and Pmode = SImode, if it is not zero_extend or ;; scalar != 1, such gather/scatter is not allowed since we don't have RVVM32SImode. (define_mode_iterator RATIO64 [ - (RVVMF8QI "TARGET_MIN_VLEN > 32") - (RVVMF4HI "TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM1DI "TARGET_VECTOR_ELEN_64") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") ]) (define_mode_iterator RATIO32 [ - RVVMF4QI - RVVMF2HI + (RVVMF4QI "!TARGET_XTHEADVECTOR") + (RVVMF2HI "!TARGET_XTHEADVECTOR") RVVM1SI (RVVM2DI "TARGET_VECTOR_ELEN_64") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") ]) (define_mode_iterator RATIO16 [ - RVVMF2QI + (RVVMF2QI "!TARGET_XTHEADVECTOR") RVVM1HI RVVM2SI (RVVM4DI "TARGET_VECTOR_ELEN_64") @@ -814,21 +814,21 @@ ]) (define_mode_iterator RATIO64I [ - (RVVMF8QI "TARGET_MIN_VLEN > 32") - (RVVMF4HI "TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM1DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") ]) (define_mode_iterator RATIO32I [ - RVVMF4QI - RVVMF2HI + (RVVMF4QI "!TARGET_XTHEADVECTOR") + (RVVMF2HI "!TARGET_XTHEADVECTOR") RVVM1SI (RVVM2DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") ]) (define_mode_iterator RATIO16I [ - RVVMF2QI + (RVVMF2QI "!TARGET_XTHEADVECTOR") RVVM1HI RVVM2SI (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") @@ -873,21 +873,21 @@ ]) (define_mode_iterator V_FRACT [ - RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") ]) (define_mode_iterator VWEXTI [ - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -933,7 +933,7 @@ (RVVM4SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -966,7 +966,7 @@ (RVVM4SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -996,7 +996,7 @@ (define_mode_iterator VWCONVERTI [ (RVVM8SI "TARGET_ZVFH") (RVVM4SI "TARGET_ZVFH") (RVVM2SI "TARGET_ZVFH") (RVVM1SI "TARGET_ZVFH") - (RVVMF2SI "TARGET_ZVFH") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_ZVFH") (RVVM8DI "TARGET_VECTOR_ELEN_64 && TARGET_VECTOR_ELEN_FP_32") (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_VECTOR_ELEN_FP_32") @@ -1045,7 +1045,7 @@ ]) (define_mode_iterator VQEXTI [ - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -1456,11 +1456,11 @@ ;; VINDEXED [VI8 VI16 VI32 (VI64 "TARGET_64BIT")]. (define_mode_iterator VINDEXED [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") @@ -1468,12 +1468,12 @@ (RVVM1DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64 && TARGET_64BIT") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64 && TARGET_64BIT") @@ -3173,11 +3173,11 @@ (define_mode_iterator V_VLS_F_CONVERT_SI [ (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") - (RVVMF2HF "TARGET_ZVFH") (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") @@ -3290,12 +3290,12 @@ ]) (define_mode_iterator V_VLS_F_CONVERT_DI [ - (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index 24b7b4394be..3eb6daafbc2 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -87,7 +87,7 @@ ;; check. However, we need default value of SEW for vsetvl instruction since there ;; is no field for ratio in the vsetvl instruction encoding. (define_attr "sew" "" - (cond [(eq_attr "mode" "RVVMF64BI,RVVMF32BI,RVVMF16BI,RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ + (cond [(eq_attr "mode" "RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ RVVM8QI,RVVM4QI,RVVM2QI,RVVM1QI,RVVMF2QI,RVVMF4QI,RVVMF8QI,\ RVVM1x8QI,RVVMF2x8QI,RVVMF4x8QI,RVVMF8x8QI,\ RVVM1x7QI,RVVMF2x7QI,RVVMF4x7QI,RVVMF8x7QI,\ @@ -99,6 +99,18 @@ V1QI,V2QI,V4QI,V8QI,V16QI,V32QI,V64QI,V128QI,V256QI,V512QI,V1024QI,V2048QI,V4096QI,\ V1BI,V2BI,V4BI,V8BI,V16BI,V32BI,V64BI,V128BI,V256BI,V512BI,V1024BI,V2048BI,V4096BI") (const_int 8) + (eq_attr "mode" "RVVMF16BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 16) + (const_int 8)) + (eq_attr "mode" "RVVMF32BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 32) + (const_int 8)) + (eq_attr "mode" "RVVMF64BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 64) + (const_int 8)) (eq_attr "mode" "RVVM8HI,RVVM4HI,RVVM2HI,RVVM1HI,RVVMF2HI,RVVMF4HI,\ RVVM1x8HI,RVVMF2x8HI,RVVMF4x8HI,\ RVVM1x7HI,RVVMF2x7HI,RVVMF4x7HI,\ @@ -159,9 +171,9 @@ (eq_attr "mode" "RVVM4QI,RVVMF2BI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2QI,RVVMF4BI") (symbol_ref "riscv_vector::LMUL_2") (eq_attr "mode" "RVVM1QI,RVVMF8BI") (symbol_ref "riscv_vector::LMUL_1") - (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "riscv_vector::LMUL_F2") - (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "riscv_vector::LMUL_F4") - (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "riscv_vector::LMUL_F8") + (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F2") + (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F4") + (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F8") (eq_attr "mode" "RVVM8HI") (symbol_ref "riscv_vector::LMUL_8") (eq_attr "mode" "RVVM4HI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2HI") (symbol_ref "riscv_vector::LMUL_2") @@ -436,6 +448,10 @@ vaeskf1,vaeskf2,vaesz,vsha2ms,vsha2ch,vsha2cl,vsm4k,vsm4r,\ vsm3me,vsm3c") (const_int INVALID_ATTRIBUTE) + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE) (eq_attr "mode" "RVVM8QI,RVVM1BI") (const_int 1) (eq_attr "mode" "RVVM4QI,RVVMF2BI") (const_int 2) (eq_attr "mode" "RVVM2QI,RVVMF4BI") (const_int 4) @@ -905,6 +921,8 @@ (symbol_ref "riscv_vector::FRM_DYN")] (symbol_ref "riscv_vector::FRM_NONE"))) +(include "thead-vector.md") + ;; ----------------------------------------------------------------- ;; ---- Miscellaneous Operations ;; ----------------------------------------------------------------- @@ -1078,6 +1096,12 @@ - We can not leave it to TARGET_SECONDARY_RELOAD since it happens before spilling. The clobber scratch is used by spilling fractional registers in IRA/LRA so it's too early. */ + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; @@ -1114,7 +1138,7 @@ (define_insn "*mov<mode>_whole" [(set (match_operand:V_WHOLE 0 "reg_or_mem_operand" "=vr, m,vr") (match_operand:V_WHOLE 1 "reg_or_mem_operand" " m,vr,vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "@ vl%m1re<sew>.v\t%0,%1 vs%m1r.v\t%1,%0 @@ -1135,6 +1159,13 @@ (match_operand:VB 1 "general_operand"))] "TARGET_VECTOR" { + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } + if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; }) @@ -1142,7 +1173,7 @@ (define_insn "*mov<mode>" [(set (match_operand:VB 0 "register_operand" "=vr") (match_operand:VB 1 "register_operand" " vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "vmv1r.v\t%0,%1" [(set_attr "type" "vmov") (set_attr "mode" "<MODE>")]) @@ -3692,7 +3723,7 @@ (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3713,7 +3744,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3734,7 +3765,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c index 2e0e12aa045..2eef9e1e1a8 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c @@ -1,4 +1,4 @@ -/* { dg-do compile } */ +/* { dg-do compile { target { ! riscv_xtheadvector } } } */ /* { dg-skip-if "test rvv intrinsic" { *-*-* } { "*" } { "-march=rv*v*" } } */ void foo0 () {__rvv_bool64_t t;} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c index 3d81b179235..ef329e30785 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c @@ -1,4 +1,4 @@ /* { dg-do compile } */ /* { dg-options "-O3 -march=rv32gc -mabi=ilp32d" } */ -#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' extension enabled} } */ +#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' or 'XTHEADVECTOR' extension enabled} } */ diff --git a/gcc/testsuite/lib/target-supports.exp b/gcc/testsuite/lib/target-supports.exp index b27c30b8c51..bf0ad5342d6 100644 --- a/gcc/testsuite/lib/target-supports.exp +++ b/gcc/testsuite/lib/target-supports.exp @@ -1956,6 +1956,18 @@ proc check_effective_target_riscv_zbb { } { }] } +# Return 1 if the target arch supports the XTheadVector extension, 0 otherwise. +# Cache the result. + +proc check_effective_target_riscv_xtheadvector { } { + return [check_no_compiler_messages riscv_ext_xtheadvector assembly { + #ifndef __riscv_xtheadvector + #error "Not __riscv_xtheadvector" + #endif + }] +} + + # Return 1 if we can execute code when using dg-add-options riscv_v proc check_effective_target_riscv_v_ok { } { -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* [PATCH v5] RISC-V: Handle differences between XTheadvector and Vector 2024-01-09 3:18 ` [PATCH v5] " Jun Sha (Joshua) 2024-01-09 22:33 ` 钟居哲 @ 2024-01-10 2:22 ` Jun Sha (Joshua) 2024-01-10 2:34 ` juzhe.zhong 2024-01-11 11:03 ` [PATCH " Jun Sha (Joshua) 1 sibling, 2 replies; 133+ messages in thread From: Jun Sha (Joshua) @ 2024-01-10 2:22 UTC (permalink / raw) To: gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, Jun Sha (Joshua), Jin Ma, Xianmiao Qu This patch is to handle the differences in instruction generation between Vector and XTheadVector. In this version, we only support partial xtheadvector instructions that leverage directly from current RVV1.0 with simple adding "th." prefix. For different name xtheadvector instructions but share same patterns as RVV1.0 instructions, we will use ASM targethook to rewrite the whole string of the instructions in the following patches. For some vector patterns that cannot be avoided, we use "!TARGET_XTHEADVECTOR" to disable them in vector.md in order not to generate instructions that xtheadvector does not support, like vmv1r and vsext.vf2. gcc/ChangeLog: * config.gcc: Add files for XTheadVector intrinsics. * config/riscv/autovec.md: Guard XTheadVector. * config/riscv/riscv-c.cc: Add pragma for XTheadVector. * config/riscv/riscv-string.cc (expand_block_move): Guard XTheadVector. * config/riscv/riscv-string.cc (vls_mode_valid_p): Avoid autovec. * config/riscv/riscv-vector-builtins-shapes.cc (check_type): (build_one): New function. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): (DEF_THEAD_RVV_FUNCTION): Add new marcos. (check_required_extensions): (handle_pragma_vector): * config/riscv/riscv-vector-builtins.h (RVV_REQUIRE_VECTOR): (RVV_REQUIRE_XTHEADVECTOR): Add RVV_REQUIRE_VECTOR and RVV_REQUIRE_XTHEADVECTOR. (struct function_group_info): * config/riscv/riscv-vector-switch.def (ENTRY): Disable fractional mode for the XTheadVector extension. (TUPLE_ENTRY): Likewise. * config/riscv/riscv-vsetvl.cc: Add functions for xtheadvector. * config/riscv/riscv.cc (riscv_v_ext_vls_mode_p): Guard XTheadVector. (riscv_v_adjust_bytesize): Likewise. (riscv_preferred_simd_mode): Likewsie. (riscv_autovectorize_vector_modes): Likewise. (riscv_vector_mode_supported_any_target_p): Likewise. (TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P): Likewise. * config/riscv/vector.md: Include thead-vector.md. * config/riscv/riscv_th_vector.h: New file. * config/riscv/thead-vector.md: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/base/pragma-1.c: Add XTheadVector. * gcc.target/riscv/rvv/base/abi-1.c: Exclude XTheadVector. * lib/target-supports.exp: Add target for XTheadVector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config.gcc | 2 +- gcc/config/riscv/autovec.md | 2 +- gcc/config/riscv/predicates.md | 4 +- gcc/config/riscv/riscv-c.cc | 3 +- gcc/config/riscv/riscv-string.cc | 3 +- gcc/config/riscv/riscv-v.cc | 2 +- .../riscv/riscv-vector-builtins-bases.cc | 48 ++++-- .../riscv/riscv-vector-builtins-shapes.cc | 23 +++ gcc/config/riscv/riscv-vector-switch.def | 150 +++++++++--------- gcc/config/riscv/riscv.cc | 20 ++- gcc/config/riscv/riscv_th_vector.h | 49 ++++++ gcc/config/riscv/thead-vector.md | 102 ++++++++++++ gcc/config/riscv/thead.cc | 23 ++- gcc/config/riscv/vector.md | 49 ++++-- .../gcc.target/riscv/rvv/base/abi-1.c | 2 +- .../gcc.target/riscv/rvv/base/pragma-1.c | 2 +- gcc/testsuite/lib/target-supports.exp | 12 ++ 17 files changed, 383 insertions(+), 113 deletions(-) create mode 100644 gcc/config/riscv/riscv_th_vector.h create mode 100644 gcc/config/riscv/thead-vector.md diff --git a/gcc/config.gcc b/gcc/config.gcc index 7e583390024..047e4c02cf4 100644 --- a/gcc/config.gcc +++ b/gcc/config.gcc @@ -549,7 +549,7 @@ riscv*) extra_objs="${extra_objs} riscv-vector-builtins.o riscv-vector-builtins-shapes.o riscv-vector-builtins-bases.o" extra_objs="${extra_objs} thead.o riscv-target-attr.o" d_target_objs="riscv-d.o" - extra_headers="riscv_vector.h" + extra_headers="riscv_vector.h riscv_th_vector.h" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.cc" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.h" ;; diff --git a/gcc/config/riscv/autovec.md b/gcc/config/riscv/autovec.md index 775eaa825b0..0477781cabe 100644 --- a/gcc/config/riscv/autovec.md +++ b/gcc/config/riscv/autovec.md @@ -2579,7 +2579,7 @@ [(match_operand 0 "register_operand") (match_operand 1 "memory_operand") (match_operand:ANYI 2 "const_int_operand")] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" { riscv_vector::expand_rawmemchr(<MODE>mode, operands[0], operands[1], operands[2]); diff --git a/gcc/config/riscv/predicates.md b/gcc/config/riscv/predicates.md index b1a79cae50a..0337da88284 100644 --- a/gcc/config/riscv/predicates.md +++ b/gcc/config/riscv/predicates.md @@ -428,7 +428,9 @@ ;; Predicates for the V extension. (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (ior (match_test "TARGET_XTHEADVECTOR && rtx_equal_p (op, const0_rtx)") + (match_test "!TARGET_XTHEADVECTOR")) + (match_operand 0 "const_csr_operand")))) (define_special_predicate "autovec_length_operand" (ior (match_operand 0 "pmode_register_operand") diff --git a/gcc/config/riscv/riscv-c.cc b/gcc/config/riscv/riscv-c.cc index 422ddc2c308..2e306057347 100644 --- a/gcc/config/riscv/riscv-c.cc +++ b/gcc/config/riscv/riscv-c.cc @@ -195,7 +195,8 @@ riscv_pragma_intrinsic (cpp_reader *) const char *name = TREE_STRING_POINTER (x); - if (strcmp (name, "vector") == 0) + if (strcmp (name, "vector") == 0 + || strcmp (name, "xtheadvector") == 0) { if (!TARGET_VECTOR) { diff --git a/gcc/config/riscv/riscv-string.cc b/gcc/config/riscv/riscv-string.cc index f227b8667ce..b09b51d7526 100644 --- a/gcc/config/riscv/riscv-string.cc +++ b/gcc/config/riscv/riscv-string.cc @@ -773,7 +773,8 @@ riscv_expand_block_move_scalar (rtx dest, rtx src, rtx length) bool riscv_expand_block_move (rtx dest, rtx src, rtx length) { - if (TARGET_VECTOR && stringop_strategy & STRATEGY_VECTOR) + if ((TARGET_VECTOR && !TARGET_XTHEADVECTOR) + && stringop_strategy & STRATEGY_VECTOR) { bool ok = riscv_vector::expand_block_move (dest, src, length); if (ok) diff --git a/gcc/config/riscv/riscv-v.cc b/gcc/config/riscv/riscv-v.cc index 2491522191a..7e25deb966d 100644 --- a/gcc/config/riscv/riscv-v.cc +++ b/gcc/config/riscv/riscv-v.cc @@ -4392,7 +4392,7 @@ cmp_lmul_gt_one (machine_mode mode) bool vls_mode_valid_p (machine_mode vls_mode) { - if (!TARGET_VECTOR) + if (!TARGET_VECTOR || TARGET_XTHEADVECTOR) return false; if (riscv_autovec_preference == RVV_SCALABLE) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 91af2c00674..46f1a1da33e 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -115,23 +115,37 @@ public: tree type = builtin_types[e.type.index].vector; machine_mode mode = TYPE_MODE (type); - /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. - - - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 - - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 - - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 - - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 - - e8,m2/e16,m4/e32,m8 --> e8m2 - - e8,m4/e16,m8 --> e8m4 - - e8,m8 --> e8m8 - */ - /* SEW. */ - e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); - - /* LMUL. */ - machine_mode e8_mode - = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); - e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + + if (TARGET_XTHEADVECTOR) + { + machine_mode inner_mode = GET_MODE_INNER (mode); + /* SEW. */ + e.add_input_operand (Pmode, + gen_int_mode (GET_MODE_BITSIZE (inner_mode), Pmode)); + /* LMUL. */ + e.add_input_operand (Pmode, + gen_int_mode (get_vlmul (mode), Pmode)); + } + else + { + /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. + + - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 + - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 + - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 + - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 + - e8,m2/e16,m4/e32,m8 --> e8m2 + - e8,m4/e16,m8 --> e8m4 + - e8,m8 --> e8m8 + */ + /* SEW. */ + e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); + + /* LMUL. */ + machine_mode e8_mode + = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); + e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + } /* TAIL_ANY. */ e.add_input_operand (Pmode, diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index ee8058dcdb1..1e4f4d53de6 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -33,6 +33,25 @@ namespace riscv_vector { +/* Check whether the RETURN_TYPE and ARGUMENT_TYPES are + valid for the function. */ + +static bool +check_type (tree return_type, vec<tree> &argument_types) +{ + tree arg; + unsigned i; + + if (!return_type) + return false; + + FOR_EACH_VEC_ELT (argument_types, i, arg) + if (!arg) + return false; + + return true; +} + /* Add one function instance for GROUP, using operand suffix at index OI, mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ static void @@ -49,6 +68,10 @@ build_one (function_builder &b, const function_group_info &group, group.ops_infos.types[vec_type_idx].index); b.allocate_argument_types (function_instance, argument_types); b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + b.add_overloaded_function (function_instance, *group.shape); b.add_unique_function (function_instance, (*group.shape), return_type, argument_types); diff --git a/gcc/config/riscv/riscv-vector-switch.def b/gcc/config/riscv/riscv-vector-switch.def index 1ad26c2a3b2..452283b7416 100644 --- a/gcc/config/riscv/riscv-vector-switch.def +++ b/gcc/config/riscv/riscv-vector-switch.def @@ -68,9 +68,9 @@ Encode the ratio of SEW/LMUL into the mask types. #endif /* Disable modes if TARGET_MIN_VLEN == 32. */ -ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) -ENTRY (RVVMF32BI, true, LMUL_F4, 32) -ENTRY (RVVMF16BI, true, LMUL_F2, 16) +ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F8, 64) +ENTRY (RVVMF32BI, true, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F4, 32) +ENTRY (RVVMF16BI, true, TARGET_XTHEADVECTOR ? LMUL_1 : LMUL_F2 , 16) ENTRY (RVVMF8BI, true, LMUL_1, 8) ENTRY (RVVMF4BI, true, LMUL_2, 4) ENTRY (RVVMF2BI, true, LMUL_4, 2) @@ -81,39 +81,39 @@ ENTRY (RVVM8QI, true, LMUL_8, 1) ENTRY (RVVM4QI, true, LMUL_4, 2) ENTRY (RVVM2QI, true, LMUL_2, 4) ENTRY (RVVM1QI, true, LMUL_1, 8) -ENTRY (RVVMF2QI, true, LMUL_F2, 16) -ENTRY (RVVMF4QI, true, LMUL_F4, 32) -ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) +ENTRY (RVVMF2QI, !TARGET_XTHEADVECTOR, LMUL_F2, 16) +ENTRY (RVVMF4QI, !TARGET_XTHEADVECTOR, LMUL_F4, 32) +ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F8, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8HI, true, LMUL_8, 2) ENTRY (RVVM4HI, true, LMUL_4, 4) ENTRY (RVVM2HI, true, LMUL_2, 8) ENTRY (RVVM1HI, true, LMUL_1, 16) -ENTRY (RVVMF2HI, true, LMUL_F2, 32) -ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HI, !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_16. */ ENTRY (RVVM8HF, TARGET_VECTOR_ELEN_FP_16, LMUL_8, 2) ENTRY (RVVM4HF, TARGET_VECTOR_ELEN_FP_16, LMUL_4, 4) ENTRY (RVVM2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_2, 8) ENTRY (RVVM1HF, TARGET_VECTOR_ELEN_FP_16, LMUL_1, 16) -ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_F2, 32) -ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8SI, true, LMUL_8, 4) ENTRY (RVVM4SI, true, LMUL_4, 8) ENTRY (RVVM2SI, true, LMUL_2, 16) ENTRY (RVVM1SI, true, LMUL_1, 32) -ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_32. */ ENTRY (RVVM8SF, TARGET_VECTOR_ELEN_FP_32, LMUL_8, 4) ENTRY (RVVM4SF, TARGET_VECTOR_ELEN_FP_32, LMUL_4, 8) ENTRY (RVVM2SF, TARGET_VECTOR_ELEN_FP_32, LMUL_2, 16) ENTRY (RVVM1SF, TARGET_VECTOR_ELEN_FP_32, LMUL_1, 32) -ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if !TARGET_VECTOR_ELEN_64. */ ENTRY (RVVM8DI, TARGET_VECTOR_ELEN_64, LMUL_8, 8) @@ -140,127 +140,127 @@ ENTRY (RVVM1DF, TARGET_VECTOR_ELEN_FP_64, LMUL_1, 64) #endif TUPLE_ENTRY (RVVM1x8QI, true, RVVM1QI, 8, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x8QI, true, RVVMF2QI, 8, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x8QI, true, RVVMF4QI, 8, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 8, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x8QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 8, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x8QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 8, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 8, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x7QI, true, RVVM1QI, 7, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x7QI, true, RVVMF2QI, 7, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x7QI, true, RVVMF4QI, 7, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 7, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x7QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 7, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x7QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 7, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 7, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x6QI, true, RVVM1QI, 6, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x6QI, true, RVVMF2QI, 6, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x6QI, true, RVVMF4QI, 6, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 6, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x6QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 6, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x6QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 6, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 6, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x5QI, true, RVVM1QI, 5, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x5QI, true, RVVMF2QI, 5, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x5QI, true, RVVMF4QI, 5, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 5, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x5QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 5, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x5QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 5, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 5, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x4QI, true, RVVM2QI, 4, LMUL_2, 4) TUPLE_ENTRY (RVVM1x4QI, true, RVVM1QI, 4, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x4QI, true, RVVMF2QI, 4, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x4QI, true, RVVMF4QI, 4, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 4, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x4QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 4, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x4QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 4, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 4, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x3QI, true, RVVM2QI, 3, LMUL_2, 4) TUPLE_ENTRY (RVVM1x3QI, true, RVVM1QI, 3, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x3QI, true, RVVMF2QI, 3, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x3QI, true, RVVMF4QI, 3, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 3, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x3QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 3, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x3QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 3, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 3, LMUL_F8, 64) TUPLE_ENTRY (RVVM4x2QI, true, RVVM4QI, 2, LMUL_4, 2) TUPLE_ENTRY (RVVM2x2QI, true, RVVM2QI, 2, LMUL_2, 4) TUPLE_ENTRY (RVVM1x2QI, true, RVVM1QI, 2, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x2QI, true, RVVMF2QI, 2, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x2QI, true, RVVMF4QI, 2, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 2, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x2QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 2, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x2QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 2, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 2, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x8HI, true, RVVM1HI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HI, true, RVVMF2HI, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HI, true, RVVM1HI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HI, true, RVVMF2HI, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HI, true, RVVM1HI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HI, true, RVVMF2HI, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HI, true, RVVM1HI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HI, true, RVVMF2HI, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HI, true, RVVM2HI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HI, true, RVVM1HI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HI, true, RVVMF2HI, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HI, true, RVVM2HI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HI, true, RVVM1HI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HI, true, RVVMF2HI, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HI, true, RVVM4HI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HI, true, RVVM2HI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HI, true, RVVM1HI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HI, true, RVVMF2HI, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM4HF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8SI, true, RVVM1SI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SI, true, RVVM1SI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SI, true, RVVM1SI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SI, true, RVVM1SI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SI, true, RVVM2SI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SI, true, RVVM1SI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SI, true, RVVM2SI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SI, true, RVVM1SI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SI, true, RVVM4SI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SI, true, RVVM2SI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SI, true, RVVM1SI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM4SF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 8, LMUL_1, 16) TUPLE_ENTRY (RVVM1x7DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 7, LMUL_1, 16) diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index 8f74cd2e502..28cf1f414e9 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -1406,6 +1406,9 @@ riscv_v_adjust_bytesize (machine_mode mode, int scale) { if (riscv_v_ext_vector_mode_p (mode)) { + if (TARGET_XTHEADVECTOR) + return BYTES_PER_RISCV_VECTOR; + poly_int64 nunits = GET_MODE_NUNITS (mode); poly_int64 mode_size = GET_MODE_SIZE (mode); @@ -9995,7 +9998,7 @@ riscv_use_divmod_expander (void) static machine_mode riscv_preferred_simd_mode (scalar_mode mode) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::preferred_simd_mode (mode); return word_mode; @@ -10346,7 +10349,7 @@ riscv_mode_priority (int, int n) unsigned int riscv_autovectorize_vector_modes (vector_modes *modes, bool all) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::autovectorize_vector_modes (modes, all); return default_autovectorize_vector_modes (modes, all); @@ -10529,6 +10532,16 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) return false; } +/* Implements target hook vector_mode_supported_any_target_p. */ + +static bool +riscv_vector_mode_supported_any_target_p (machine_mode mode) +{ + if (TARGET_XTHEADVECTOR) + return false; + return true; +} + /* Initialize the GCC target structure. */ #undef TARGET_ASM_ALIGNED_HI_OP #define TARGET_ASM_ALIGNED_HI_OP "\t.half\t" @@ -10872,6 +10885,9 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) #undef TARGET_PREFERRED_ELSE_VALUE #define TARGET_PREFERRED_ELSE_VALUE riscv_preferred_else_value +#undef TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P +#define TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P riscv_vector_mode_supported_any_target_p + struct gcc_target targetm = TARGET_INITIALIZER; #include "gt-riscv.h" diff --git a/gcc/config/riscv/riscv_th_vector.h b/gcc/config/riscv/riscv_th_vector.h new file mode 100644 index 00000000000..b6b6738bdda --- /dev/null +++ b/gcc/config/riscv/riscv_th_vector.h @@ -0,0 +1,49 @@ +/* RISC-V 'XTheadVector' Extension intrinsics include file. + Copyright (C) 2024 Free Software Foundation, Inc. + + This file is part of GCC. + + GCC is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published + by the Free Software Foundation; either version 3, or (at your + option) any later version. + + GCC is distributed in the hope that it will be useful, but WITHOUT + ANY WARRANTY; without even the implied warranty of MERCHANTABILITY + or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public + License for more details. + + Under Section 7 of GPL version 3, you are granted additional + permissions described in the GCC Runtime Library Exception, version + 3.1, as published by the Free Software Foundation. + + You should have received a copy of the GNU General Public License and + a copy of the GCC Runtime Library Exception along with this program; + see the files COPYING3 and COPYING.RUNTIME respectively. If not, see + <http://www.gnu.org/licenses/>. */ + +#ifndef __RISCV_TH_VECTOR_H +#define __RISCV_TH_VECTOR_H + +#include <stdint.h> +#include <stddef.h> + +#ifndef __riscv_xtheadvector +#error "XTheadVector intrinsics require the xtheadvector extension." +#else +#ifdef __cplusplus +extern "C" { +#endif + +/* NOTE: This implementation of riscv_th_vector.h is intentionally short. It does + not define the RVV types and intrinsic functions directly in C and C++ + code, but instead uses the following pragma to tell GCC to insert the + necessary type and function definitions itself. The net effect is the + same, and the file is a complete implementation of riscv_th_vector.h. */ +#pragma riscv intrinsic "xtheadvector" + +#ifdef __cplusplus +} +#endif // __cplusplus +#endif // __riscv_xtheadvector +#endif // __RISCV_TH_ECTOR_H diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md new file mode 100644 index 00000000000..696b815252d --- /dev/null +++ b/gcc/config/riscv/thead-vector.md @@ -0,0 +1,102 @@ +(define_c_enum "unspec" [ + UNSPEC_TH_VWLDST +]) + +(define_mode_iterator V_VLS_VT [V VLS VT]) +(define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) + +(define_split + [(set (match_operand:V_VB_VLS_VT 0 "reg_or_mem_operand") + (match_operand:V_VB_VLS_VT 1 "reg_or_mem_operand"))] + "TARGET_XTHEADVECTOR" + [(const_int 0)] + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + }) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:V_VLS_VT 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:V_VLS_VT + [(match_operand:V_VLS_VT 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2")]) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:VB 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:VB + [(match_operand:VB 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2") + (set (attr "sew") (const_int 8)) + (set (attr "vlmul") (symbol_ref "riscv_vector::LMUL_1"))]) + +(define_insn_and_split "*pred_th_mov<mode>" + [(set (match_operand:VB_VLS 0 "nonimmediate_operand" "=vr, m, vr, vr, vr") + (if_then_else:VB_VLS + (unspec:VB_VLS + [(match_operand:VB_VLS 1 "vector_all_trues_mask_operand" "Wc1, Wc1, Wc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) + (match_operand:VB_VLS 3 "vector_move_operand" " m, vr, vr, Wc0, Wc1") + (match_operand:VB_VLS 2 "vector_undef_operand" " vu, vu, vu, vu, vu")))] + "TARGET_XTHEADVECTOR" + "@ + # + # + vmcpy.m\t%0,%3 + vmclr.m\t%0 + vmset.m\t%0" + "&& !reload_completed" + [(const_int 0)] + { + if ((MEM_P (operands[0]) || MEM_P (operands[3])) + || (REG_P (operands[0]) && REG_P (operands[3]) + && INTVAL (operands[5]) == riscv_vector::VLMAX)) + { + emit_move_insn (operands[0], operands[3]); + DONE; + } + FAIL; + } + [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/config/riscv/thead.cc b/gcc/config/riscv/thead.cc index 00b2e14c6c0..cc04101b256 100644 --- a/gcc/config/riscv/thead.cc +++ b/gcc/config/riscv/thead.cc @@ -27,6 +27,7 @@ #include "backend.h" #include "tree.h" #include "rtl.h" +#include "insn-attr.h" #include "explow.h" #include "memmodel.h" #include "emit-rtl.h" @@ -890,8 +891,26 @@ th_asm_output_opcode (FILE *asm_out_file, const char *p) { /* We need to add th. prefix to all the xtheadvector instructions here.*/ - if (current_output_insn != NULL && p[0] == 'v') - fputs ("th.", asm_out_file); + if (current_output_insn != NULL) + { + if (get_attr_type (current_output_insn) == TYPE_VSETVL) + { + if (strstr (p, "zero")) + { + if (strstr (p, "zero,zero")) + return "th.vsetvli\tzero,zero,e%0,%m1"; + else + return "th.vsetvli\tzero,%0,e%1,%m2"; + } + else + { + return "th.vsetvli\t%0,%1,e%2,%m3"; + } + } + + if (p[0] == 'v') + fputs ("th.", asm_out_file); + } return p; } diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index 24b7b4394be..3eb6daafbc2 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -87,7 +87,7 @@ ;; check. However, we need default value of SEW for vsetvl instruction since there ;; is no field for ratio in the vsetvl instruction encoding. (define_attr "sew" "" - (cond [(eq_attr "mode" "RVVMF64BI,RVVMF32BI,RVVMF16BI,RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ + (cond [(eq_attr "mode" "RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ RVVM8QI,RVVM4QI,RVVM2QI,RVVM1QI,RVVMF2QI,RVVMF4QI,RVVMF8QI,\ RVVM1x8QI,RVVMF2x8QI,RVVMF4x8QI,RVVMF8x8QI,\ RVVM1x7QI,RVVMF2x7QI,RVVMF4x7QI,RVVMF8x7QI,\ @@ -99,6 +99,18 @@ V1QI,V2QI,V4QI,V8QI,V16QI,V32QI,V64QI,V128QI,V256QI,V512QI,V1024QI,V2048QI,V4096QI,\ V1BI,V2BI,V4BI,V8BI,V16BI,V32BI,V64BI,V128BI,V256BI,V512BI,V1024BI,V2048BI,V4096BI") (const_int 8) + (eq_attr "mode" "RVVMF16BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 16) + (const_int 8)) + (eq_attr "mode" "RVVMF32BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 32) + (const_int 8)) + (eq_attr "mode" "RVVMF64BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 64) + (const_int 8)) (eq_attr "mode" "RVVM8HI,RVVM4HI,RVVM2HI,RVVM1HI,RVVMF2HI,RVVMF4HI,\ RVVM1x8HI,RVVMF2x8HI,RVVMF4x8HI,\ RVVM1x7HI,RVVMF2x7HI,RVVMF4x7HI,\ @@ -159,9 +171,9 @@ (eq_attr "mode" "RVVM4QI,RVVMF2BI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2QI,RVVMF4BI") (symbol_ref "riscv_vector::LMUL_2") (eq_attr "mode" "RVVM1QI,RVVMF8BI") (symbol_ref "riscv_vector::LMUL_1") - (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "riscv_vector::LMUL_F2") - (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "riscv_vector::LMUL_F4") - (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "riscv_vector::LMUL_F8") + (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F2") + (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F4") + (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F8") (eq_attr "mode" "RVVM8HI") (symbol_ref "riscv_vector::LMUL_8") (eq_attr "mode" "RVVM4HI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2HI") (symbol_ref "riscv_vector::LMUL_2") @@ -436,6 +448,10 @@ vaeskf1,vaeskf2,vaesz,vsha2ms,vsha2ch,vsha2cl,vsm4k,vsm4r,\ vsm3me,vsm3c") (const_int INVALID_ATTRIBUTE) + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE) (eq_attr "mode" "RVVM8QI,RVVM1BI") (const_int 1) (eq_attr "mode" "RVVM4QI,RVVMF2BI") (const_int 2) (eq_attr "mode" "RVVM2QI,RVVMF4BI") (const_int 4) @@ -905,6 +921,8 @@ (symbol_ref "riscv_vector::FRM_DYN")] (symbol_ref "riscv_vector::FRM_NONE"))) +(include "thead-vector.md") + ;; ----------------------------------------------------------------- ;; ---- Miscellaneous Operations ;; ----------------------------------------------------------------- @@ -1078,6 +1096,12 @@ - We can not leave it to TARGET_SECONDARY_RELOAD since it happens before spilling. The clobber scratch is used by spilling fractional registers in IRA/LRA so it's too early. */ + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; @@ -1114,7 +1138,7 @@ (define_insn "*mov<mode>_whole" [(set (match_operand:V_WHOLE 0 "reg_or_mem_operand" "=vr, m,vr") (match_operand:V_WHOLE 1 "reg_or_mem_operand" " m,vr,vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "@ vl%m1re<sew>.v\t%0,%1 vs%m1r.v\t%1,%0 @@ -1135,6 +1159,13 @@ (match_operand:VB 1 "general_operand"))] "TARGET_VECTOR" { + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } + if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; }) @@ -1142,7 +1173,7 @@ (define_insn "*mov<mode>" [(set (match_operand:VB 0 "register_operand" "=vr") (match_operand:VB 1 "register_operand" " vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "vmv1r.v\t%0,%1" [(set_attr "type" "vmov") (set_attr "mode" "<MODE>")]) @@ -3692,7 +3723,7 @@ (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3713,7 +3744,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3734,7 +3765,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c index 2e0e12aa045..2eef9e1e1a8 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c @@ -1,4 +1,4 @@ -/* { dg-do compile } */ +/* { dg-do compile { target { ! riscv_xtheadvector } } } */ /* { dg-skip-if "test rvv intrinsic" { *-*-* } { "*" } { "-march=rv*v*" } } */ void foo0 () {__rvv_bool64_t t;} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c index 3d81b179235..ef329e30785 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c @@ -1,4 +1,4 @@ /* { dg-do compile } */ /* { dg-options "-O3 -march=rv32gc -mabi=ilp32d" } */ -#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' extension enabled} } */ +#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' or 'XTHEADVECTOR' extension enabled} } */ diff --git a/gcc/testsuite/lib/target-supports.exp b/gcc/testsuite/lib/target-supports.exp index b27c30b8c51..bf0ad5342d6 100644 --- a/gcc/testsuite/lib/target-supports.exp +++ b/gcc/testsuite/lib/target-supports.exp @@ -1956,6 +1956,18 @@ proc check_effective_target_riscv_zbb { } { }] } +# Return 1 if the target arch supports the XTheadVector extension, 0 otherwise. +# Cache the result. + +proc check_effective_target_riscv_xtheadvector { } { + return [check_no_compiler_messages riscv_ext_xtheadvector assembly { + #ifndef __riscv_xtheadvector + #error "Not __riscv_xtheadvector" + #endif + }] +} + + # Return 1 if we can execute code when using dg-add-options riscv_v proc check_effective_target_riscv_v_ok { } { -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: [PATCH v5] RISC-V: Handle differences between XTheadvector and Vector 2024-01-10 2:22 ` Jun Sha (Joshua) @ 2024-01-10 2:34 ` juzhe.zhong 2024-01-10 2:57 ` Re：[PATCH " joshua 2024-01-11 11:03 ` [PATCH " Jun Sha (Joshua) 1 sibling, 1 reply; 133+ messages in thread From: juzhe.zhong @ 2024-01-10 2:34 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, cooper.joshua, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 46178 bytes --] Thanks for your patience. LGTM from myside. I think it's pretty clean now. I can image in the future when some day the theadvector is no longer used, we can remove it very easily. And also, the theadvector won't affect our RVV1.0 maintain since it's isolated cleanly. But I'd like to wait for a few more days some body want to chime in. And you should do more things before commit it: 1. Remember you should run the full coverage RVV1.0 API test, the test-generator is downloaded from official intrinsic doc: https://github.com/riscv-non-isa/rvv-intrinsic-doc 2. Also the regression of RV32 an RV64 of GCC testsuite. Do you have more patches of theadvector that I didn't review ? plz point them to me again. Thanks. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-10 10:22 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Handle differences between XTheadvector and Vector This patch is to handle the differences in instruction generation between Vector and XTheadVector. In this version, we only support partial xtheadvector instructions that leverage directly from current RVV1.0 with simple adding "th." prefix. For different name xtheadvector instructions but share same patterns as RVV1.0 instructions, we will use ASM targethook to rewrite the whole string of the instructions in the following patches. For some vector patterns that cannot be avoided, we use "!TARGET_XTHEADVECTOR" to disable them in vector.md in order not to generate instructions that xtheadvector does not support, like vmv1r and vsext.vf2. gcc/ChangeLog: * config.gcc: Add files for XTheadVector intrinsics. * config/riscv/autovec.md: Guard XTheadVector. * config/riscv/riscv-c.cc: Add pragma for XTheadVector. * config/riscv/riscv-string.cc (expand_block_move): Guard XTheadVector. * config/riscv/riscv-string.cc (vls_mode_valid_p): Avoid autovec. * config/riscv/riscv-vector-builtins-shapes.cc (check_type): (build_one): New function. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): (DEF_THEAD_RVV_FUNCTION): Add new marcos. (check_required_extensions): (handle_pragma_vector): * config/riscv/riscv-vector-builtins.h (RVV_REQUIRE_VECTOR): (RVV_REQUIRE_XTHEADVECTOR): Add RVV_REQUIRE_VECTOR and RVV_REQUIRE_XTHEADVECTOR. (struct function_group_info): * config/riscv/riscv-vector-switch.def (ENTRY): Disable fractional mode for the XTheadVector extension. (TUPLE_ENTRY): Likewise. * config/riscv/riscv-vsetvl.cc: Add functions for xtheadvector. * config/riscv/riscv.cc (riscv_v_ext_vls_mode_p): Guard XTheadVector. (riscv_v_adjust_bytesize): Likewise. (riscv_preferred_simd_mode): Likewsie. (riscv_autovectorize_vector_modes): Likewise. (riscv_vector_mode_supported_any_target_p): Likewise. (TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P): Likewise. * config/riscv/vector.md: Include thead-vector.md. * config/riscv/riscv_th_vector.h: New file. * config/riscv/thead-vector.md: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/base/pragma-1.c: Add XTheadVector. * gcc.target/riscv/rvv/base/abi-1.c: Exclude XTheadVector. * lib/target-supports.exp: Add target for XTheadVector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config.gcc | 2 +- gcc/config/riscv/autovec.md | 2 +- gcc/config/riscv/predicates.md | 4 +- gcc/config/riscv/riscv-c.cc | 3 +- gcc/config/riscv/riscv-string.cc | 3 +- gcc/config/riscv/riscv-v.cc | 2 +- .../riscv/riscv-vector-builtins-bases.cc | 48 ++++-- .../riscv/riscv-vector-builtins-shapes.cc | 23 +++ gcc/config/riscv/riscv-vector-switch.def | 150 +++++++++--------- gcc/config/riscv/riscv.cc | 20 ++- gcc/config/riscv/riscv_th_vector.h | 49 ++++++ gcc/config/riscv/thead-vector.md | 102 ++++++++++++ gcc/config/riscv/thead.cc | 23 ++- gcc/config/riscv/vector.md | 49 ++++-- .../gcc.target/riscv/rvv/base/abi-1.c | 2 +- .../gcc.target/riscv/rvv/base/pragma-1.c | 2 +- gcc/testsuite/lib/target-supports.exp | 12 ++ 17 files changed, 383 insertions(+), 113 deletions(-) create mode 100644 gcc/config/riscv/riscv_th_vector.h create mode 100644 gcc/config/riscv/thead-vector.md diff --git a/gcc/config.gcc b/gcc/config.gcc index 7e583390024..047e4c02cf4 100644 --- a/gcc/config.gcc +++ b/gcc/config.gcc @@ -549,7 +549,7 @@ riscv*) extra_objs="${extra_objs} riscv-vector-builtins.o riscv-vector-builtins-shapes.o riscv-vector-builtins-bases.o" extra_objs="${extra_objs} thead.o riscv-target-attr.o" d_target_objs="riscv-d.o" - extra_headers="riscv_vector.h" + extra_headers="riscv_vector.h riscv_th_vector.h" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.cc" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.h" ;; diff --git a/gcc/config/riscv/autovec.md b/gcc/config/riscv/autovec.md index 775eaa825b0..0477781cabe 100644 --- a/gcc/config/riscv/autovec.md +++ b/gcc/config/riscv/autovec.md @@ -2579,7 +2579,7 @@ [(match_operand 0 "register_operand") (match_operand 1 "memory_operand") (match_operand:ANYI 2 "const_int_operand")] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" { riscv_vector::expand_rawmemchr(<MODE>mode, operands[0], operands[1], operands[2]); diff --git a/gcc/config/riscv/predicates.md b/gcc/config/riscv/predicates.md index b1a79cae50a..0337da88284 100644 --- a/gcc/config/riscv/predicates.md +++ b/gcc/config/riscv/predicates.md @@ -428,7 +428,9 @@ ;; Predicates for the V extension. (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (ior (match_test "TARGET_XTHEADVECTOR && rtx_equal_p (op, const0_rtx)") + (match_test "!TARGET_XTHEADVECTOR")) + (match_operand 0 "const_csr_operand")))) (define_special_predicate "autovec_length_operand" (ior (match_operand 0 "pmode_register_operand") diff --git a/gcc/config/riscv/riscv-c.cc b/gcc/config/riscv/riscv-c.cc index 422ddc2c308..2e306057347 100644 --- a/gcc/config/riscv/riscv-c.cc +++ b/gcc/config/riscv/riscv-c.cc @@ -195,7 +195,8 @@ riscv_pragma_intrinsic (cpp_reader *) const char *name = TREE_STRING_POINTER (x); - if (strcmp (name, "vector") == 0) + if (strcmp (name, "vector") == 0 + || strcmp (name, "xtheadvector") == 0) { if (!TARGET_VECTOR) { diff --git a/gcc/config/riscv/riscv-string.cc b/gcc/config/riscv/riscv-string.cc index f227b8667ce..b09b51d7526 100644 --- a/gcc/config/riscv/riscv-string.cc +++ b/gcc/config/riscv/riscv-string.cc @@ -773,7 +773,8 @@ riscv_expand_block_move_scalar (rtx dest, rtx src, rtx length) bool riscv_expand_block_move (rtx dest, rtx src, rtx length) { - if (TARGET_VECTOR && stringop_strategy & STRATEGY_VECTOR) + if ((TARGET_VECTOR && !TARGET_XTHEADVECTOR) + && stringop_strategy & STRATEGY_VECTOR) { bool ok = riscv_vector::expand_block_move (dest, src, length); if (ok) diff --git a/gcc/config/riscv/riscv-v.cc b/gcc/config/riscv/riscv-v.cc index 2491522191a..7e25deb966d 100644 --- a/gcc/config/riscv/riscv-v.cc +++ b/gcc/config/riscv/riscv-v.cc @@ -4392,7 +4392,7 @@ cmp_lmul_gt_one (machine_mode mode) bool vls_mode_valid_p (machine_mode vls_mode) { - if (!TARGET_VECTOR) + if (!TARGET_VECTOR || TARGET_XTHEADVECTOR) return false; if (riscv_autovec_preference == RVV_SCALABLE) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 91af2c00674..46f1a1da33e 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -115,23 +115,37 @@ public: tree type = builtin_types[e.type.index].vector; machine_mode mode = TYPE_MODE (type); - /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. - - - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 - - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 - - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 - - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 - - e8,m2/e16,m4/e32,m8 --> e8m2 - - e8,m4/e16,m8 --> e8m4 - - e8,m8 --> e8m8 - */ - /* SEW. */ - e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); - - /* LMUL. */ - machine_mode e8_mode - = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); - e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + + if (TARGET_XTHEADVECTOR) + { + machine_mode inner_mode = GET_MODE_INNER (mode); + /* SEW. */ + e.add_input_operand (Pmode, + gen_int_mode (GET_MODE_BITSIZE (inner_mode), Pmode)); + /* LMUL. */ + e.add_input_operand (Pmode, + gen_int_mode (get_vlmul (mode), Pmode)); + } + else + { + /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. + + - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 + - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 + - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 + - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 + - e8,m2/e16,m4/e32,m8 --> e8m2 + - e8,m4/e16,m8 --> e8m4 + - e8,m8 --> e8m8 + */ + /* SEW. */ + e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); + + /* LMUL. */ + machine_mode e8_mode + = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); + e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + } /* TAIL_ANY. */ e.add_input_operand (Pmode, diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index ee8058dcdb1..1e4f4d53de6 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -33,6 +33,25 @@ namespace riscv_vector { +/* Check whether the RETURN_TYPE and ARGUMENT_TYPES are + valid for the function. */ + +static bool +check_type (tree return_type, vec<tree> &argument_types) +{ + tree arg; + unsigned i; + + if (!return_type) + return false; + + FOR_EACH_VEC_ELT (argument_types, i, arg) + if (!arg) + return false; + + return true; +} + /* Add one function instance for GROUP, using operand suffix at index OI, mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ static void @@ -49,6 +68,10 @@ build_one (function_builder &b, const function_group_info &group, group.ops_infos.types[vec_type_idx].index); b.allocate_argument_types (function_instance, argument_types); b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + b.add_overloaded_function (function_instance, *group.shape); b.add_unique_function (function_instance, (*group.shape), return_type, argument_types); diff --git a/gcc/config/riscv/riscv-vector-switch.def b/gcc/config/riscv/riscv-vector-switch.def index 1ad26c2a3b2..452283b7416 100644 --- a/gcc/config/riscv/riscv-vector-switch.def +++ b/gcc/config/riscv/riscv-vector-switch.def @@ -68,9 +68,9 @@ Encode the ratio of SEW/LMUL into the mask types. #endif /* Disable modes if TARGET_MIN_VLEN == 32. */ -ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) -ENTRY (RVVMF32BI, true, LMUL_F4, 32) -ENTRY (RVVMF16BI, true, LMUL_F2, 16) +ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F8, 64) +ENTRY (RVVMF32BI, true, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F4, 32) +ENTRY (RVVMF16BI, true, TARGET_XTHEADVECTOR ? LMUL_1 : LMUL_F2 , 16) ENTRY (RVVMF8BI, true, LMUL_1, 8) ENTRY (RVVMF4BI, true, LMUL_2, 4) ENTRY (RVVMF2BI, true, LMUL_4, 2) @@ -81,39 +81,39 @@ ENTRY (RVVM8QI, true, LMUL_8, 1) ENTRY (RVVM4QI, true, LMUL_4, 2) ENTRY (RVVM2QI, true, LMUL_2, 4) ENTRY (RVVM1QI, true, LMUL_1, 8) -ENTRY (RVVMF2QI, true, LMUL_F2, 16) -ENTRY (RVVMF4QI, true, LMUL_F4, 32) -ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) +ENTRY (RVVMF2QI, !TARGET_XTHEADVECTOR, LMUL_F2, 16) +ENTRY (RVVMF4QI, !TARGET_XTHEADVECTOR, LMUL_F4, 32) +ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F8, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8HI, true, LMUL_8, 2) ENTRY (RVVM4HI, true, LMUL_4, 4) ENTRY (RVVM2HI, true, LMUL_2, 8) ENTRY (RVVM1HI, true, LMUL_1, 16) -ENTRY (RVVMF2HI, true, LMUL_F2, 32) -ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HI, !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_16. */ ENTRY (RVVM8HF, TARGET_VECTOR_ELEN_FP_16, LMUL_8, 2) ENTRY (RVVM4HF, TARGET_VECTOR_ELEN_FP_16, LMUL_4, 4) ENTRY (RVVM2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_2, 8) ENTRY (RVVM1HF, TARGET_VECTOR_ELEN_FP_16, LMUL_1, 16) -ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_F2, 32) -ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8SI, true, LMUL_8, 4) ENTRY (RVVM4SI, true, LMUL_4, 8) ENTRY (RVVM2SI, true, LMUL_2, 16) ENTRY (RVVM1SI, true, LMUL_1, 32) -ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_32. */ ENTRY (RVVM8SF, TARGET_VECTOR_ELEN_FP_32, LMUL_8, 4) ENTRY (RVVM4SF, TARGET_VECTOR_ELEN_FP_32, LMUL_4, 8) ENTRY (RVVM2SF, TARGET_VECTOR_ELEN_FP_32, LMUL_2, 16) ENTRY (RVVM1SF, TARGET_VECTOR_ELEN_FP_32, LMUL_1, 32) -ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if !TARGET_VECTOR_ELEN_64. */ ENTRY (RVVM8DI, TARGET_VECTOR_ELEN_64, LMUL_8, 8) @@ -140,127 +140,127 @@ ENTRY (RVVM1DF, TARGET_VECTOR_ELEN_FP_64, LMUL_1, 64) #endif TUPLE_ENTRY (RVVM1x8QI, true, RVVM1QI, 8, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x8QI, true, RVVMF2QI, 8, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x8QI, true, RVVMF4QI, 8, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 8, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x8QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 8, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x8QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 8, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 8, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x7QI, true, RVVM1QI, 7, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x7QI, true, RVVMF2QI, 7, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x7QI, true, RVVMF4QI, 7, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 7, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x7QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 7, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x7QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 7, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 7, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x6QI, true, RVVM1QI, 6, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x6QI, true, RVVMF2QI, 6, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x6QI, true, RVVMF4QI, 6, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 6, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x6QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 6, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x6QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 6, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 6, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x5QI, true, RVVM1QI, 5, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x5QI, true, RVVMF2QI, 5, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x5QI, true, RVVMF4QI, 5, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 5, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x5QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 5, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x5QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 5, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 5, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x4QI, true, RVVM2QI, 4, LMUL_2, 4) TUPLE_ENTRY (RVVM1x4QI, true, RVVM1QI, 4, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x4QI, true, RVVMF2QI, 4, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x4QI, true, RVVMF4QI, 4, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 4, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x4QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 4, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x4QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 4, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 4, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x3QI, true, RVVM2QI, 3, LMUL_2, 4) TUPLE_ENTRY (RVVM1x3QI, true, RVVM1QI, 3, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x3QI, true, RVVMF2QI, 3, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x3QI, true, RVVMF4QI, 3, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 3, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x3QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 3, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x3QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 3, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 3, LMUL_F8, 64) TUPLE_ENTRY (RVVM4x2QI, true, RVVM4QI, 2, LMUL_4, 2) TUPLE_ENTRY (RVVM2x2QI, true, RVVM2QI, 2, LMUL_2, 4) TUPLE_ENTRY (RVVM1x2QI, true, RVVM1QI, 2, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x2QI, true, RVVMF2QI, 2, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x2QI, true, RVVMF4QI, 2, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 2, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x2QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 2, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x2QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 2, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 2, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x8HI, true, RVVM1HI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HI, true, RVVMF2HI, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HI, true, RVVM1HI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HI, true, RVVMF2HI, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HI, true, RVVM1HI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HI, true, RVVMF2HI, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HI, true, RVVM1HI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HI, true, RVVMF2HI, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HI, true, RVVM2HI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HI, true, RVVM1HI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HI, true, RVVMF2HI, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HI, true, RVVM2HI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HI, true, RVVM1HI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HI, true, RVVMF2HI, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HI, true, RVVM4HI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HI, true, RVVM2HI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HI, true, RVVM1HI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HI, true, RVVMF2HI, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM4HF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8SI, true, RVVM1SI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SI, true, RVVM1SI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SI, true, RVVM1SI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SI, true, RVVM1SI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SI, true, RVVM2SI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SI, true, RVVM1SI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SI, true, RVVM2SI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SI, true, RVVM1SI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SI, true, RVVM4SI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SI, true, RVVM2SI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SI, true, RVVM1SI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM4SF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 8, LMUL_1, 16) TUPLE_ENTRY (RVVM1x7DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 7, LMUL_1, 16) diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index 8f74cd2e502..28cf1f414e9 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -1406,6 +1406,9 @@ riscv_v_adjust_bytesize (machine_mode mode, int scale) { if (riscv_v_ext_vector_mode_p (mode)) { + if (TARGET_XTHEADVECTOR) + return BYTES_PER_RISCV_VECTOR; + poly_int64 nunits = GET_MODE_NUNITS (mode); poly_int64 mode_size = GET_MODE_SIZE (mode); @@ -9995,7 +9998,7 @@ riscv_use_divmod_expander (void) static machine_mode riscv_preferred_simd_mode (scalar_mode mode) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::preferred_simd_mode (mode); return word_mode; @@ -10346,7 +10349,7 @@ riscv_mode_priority (int, int n) unsigned int riscv_autovectorize_vector_modes (vector_modes *modes, bool all) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::autovectorize_vector_modes (modes, all); return default_autovectorize_vector_modes (modes, all); @@ -10529,6 +10532,16 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) return false; } +/* Implements target hook vector_mode_supported_any_target_p. */ + +static bool +riscv_vector_mode_supported_any_target_p (machine_mode mode) +{ + if (TARGET_XTHEADVECTOR) + return false; + return true; +} + /* Initialize the GCC target structure. */ #undef TARGET_ASM_ALIGNED_HI_OP #define TARGET_ASM_ALIGNED_HI_OP "\t.half\t" @@ -10872,6 +10885,9 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) #undef TARGET_PREFERRED_ELSE_VALUE #define TARGET_PREFERRED_ELSE_VALUE riscv_preferred_else_value +#undef TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P +#define TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P riscv_vector_mode_supported_any_target_p + struct gcc_target targetm = TARGET_INITIALIZER; #include "gt-riscv.h" diff --git a/gcc/config/riscv/riscv_th_vector.h b/gcc/config/riscv/riscv_th_vector.h new file mode 100644 index 00000000000..b6b6738bdda --- /dev/null +++ b/gcc/config/riscv/riscv_th_vector.h @@ -0,0 +1,49 @@ +/* RISC-V 'XTheadVector' Extension intrinsics include file. + Copyright (C) 2024 Free Software Foundation, Inc. + + This file is part of GCC. + + GCC is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published + by the Free Software Foundation; either version 3, or (at your + option) any later version. + + GCC is distributed in the hope that it will be useful, but WITHOUT + ANY WARRANTY; without even the implied warranty of MERCHANTABILITY + or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public + License for more details. + + Under Section 7 of GPL version 3, you are granted additional + permissions described in the GCC Runtime Library Exception, version + 3.1, as published by the Free Software Foundation. + + You should have received a copy of the GNU General Public License and + a copy of the GCC Runtime Library Exception along with this program; + see the files COPYING3 and COPYING.RUNTIME respectively. If not, see + <http://www.gnu.org/licenses/>. */ + +#ifndef __RISCV_TH_VECTOR_H +#define __RISCV_TH_VECTOR_H + +#include <stdint.h> +#include <stddef.h> + +#ifndef __riscv_xtheadvector +#error "XTheadVector intrinsics require the xtheadvector extension." +#else +#ifdef __cplusplus +extern "C" { +#endif + +/* NOTE: This implementation of riscv_th_vector.h is intentionally short. It does + not define the RVV types and intrinsic functions directly in C and C++ + code, but instead uses the following pragma to tell GCC to insert the + necessary type and function definitions itself. The net effect is the + same, and the file is a complete implementation of riscv_th_vector.h. */ +#pragma riscv intrinsic "xtheadvector" + +#ifdef __cplusplus +} +#endif // __cplusplus +#endif // __riscv_xtheadvector +#endif // __RISCV_TH_ECTOR_H diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md new file mode 100644 index 00000000000..696b815252d --- /dev/null +++ b/gcc/config/riscv/thead-vector.md @@ -0,0 +1,102 @@ +(define_c_enum "unspec" [ + UNSPEC_TH_VWLDST +]) + +(define_mode_iterator V_VLS_VT [V VLS VT]) +(define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) + +(define_split + [(set (match_operand:V_VB_VLS_VT 0 "reg_or_mem_operand") + (match_operand:V_VB_VLS_VT 1 "reg_or_mem_operand"))] + "TARGET_XTHEADVECTOR" + [(const_int 0)] + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + }) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:V_VLS_VT 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:V_VLS_VT + [(match_operand:V_VLS_VT 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2")]) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:VB 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:VB + [(match_operand:VB 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2") + (set (attr "sew") (const_int 8)) + (set (attr "vlmul") (symbol_ref "riscv_vector::LMUL_1"))]) + +(define_insn_and_split "*pred_th_mov<mode>" + [(set (match_operand:VB_VLS 0 "nonimmediate_operand" "=vr, m, vr, vr, vr") + (if_then_else:VB_VLS + (unspec:VB_VLS + [(match_operand:VB_VLS 1 "vector_all_trues_mask_operand" "Wc1, Wc1, Wc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) + (match_operand:VB_VLS 3 "vector_move_operand" " m, vr, vr, Wc0, Wc1") + (match_operand:VB_VLS 2 "vector_undef_operand" " vu, vu, vu, vu, vu")))] + "TARGET_XTHEADVECTOR" + "@ + # + # + vmcpy.m\t%0,%3 + vmclr.m\t%0 + vmset.m\t%0" + "&& !reload_completed" + [(const_int 0)] + { + if ((MEM_P (operands[0]) || MEM_P (operands[3])) + || (REG_P (operands[0]) && REG_P (operands[3]) + && INTVAL (operands[5]) == riscv_vector::VLMAX)) + { + emit_move_insn (operands[0], operands[3]); + DONE; + } + FAIL; + } + [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/config/riscv/thead.cc b/gcc/config/riscv/thead.cc index 00b2e14c6c0..cc04101b256 100644 --- a/gcc/config/riscv/thead.cc +++ b/gcc/config/riscv/thead.cc @@ -27,6 +27,7 @@ #include "backend.h" #include "tree.h" #include "rtl.h" +#include "insn-attr.h" #include "explow.h" #include "memmodel.h" #include "emit-rtl.h" @@ -890,8 +891,26 @@ th_asm_output_opcode (FILE *asm_out_file, const char *p) { /* We need to add th. prefix to all the xtheadvector instructions here.*/ - if (current_output_insn != NULL && p[0] == 'v') - fputs ("th.", asm_out_file); + if (current_output_insn != NULL) + { + if (get_attr_type (current_output_insn) == TYPE_VSETVL) + { + if (strstr (p, "zero")) + { + if (strstr (p, "zero,zero")) + return "th.vsetvli\tzero,zero,e%0,%m1"; + else + return "th.vsetvli\tzero,%0,e%1,%m2"; + } + else + { + return "th.vsetvli\t%0,%1,e%2,%m3"; + } + } + + if (p[0] == 'v') + fputs ("th.", asm_out_file); + } return p; } diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index 24b7b4394be..3eb6daafbc2 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -87,7 +87,7 @@ ;; check. However, we need default value of SEW for vsetvl instruction since there ;; is no field for ratio in the vsetvl instruction encoding. (define_attr "sew" "" - (cond [(eq_attr "mode" "RVVMF64BI,RVVMF32BI,RVVMF16BI,RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ + (cond [(eq_attr "mode" "RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ RVVM8QI,RVVM4QI,RVVM2QI,RVVM1QI,RVVMF2QI,RVVMF4QI,RVVMF8QI,\ RVVM1x8QI,RVVMF2x8QI,RVVMF4x8QI,RVVMF8x8QI,\ RVVM1x7QI,RVVMF2x7QI,RVVMF4x7QI,RVVMF8x7QI,\ @@ -99,6 +99,18 @@ V1QI,V2QI,V4QI,V8QI,V16QI,V32QI,V64QI,V128QI,V256QI,V512QI,V1024QI,V2048QI,V4096QI,\ V1BI,V2BI,V4BI,V8BI,V16BI,V32BI,V64BI,V128BI,V256BI,V512BI,V1024BI,V2048BI,V4096BI") (const_int 8) + (eq_attr "mode" "RVVMF16BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 16) + (const_int 8)) + (eq_attr "mode" "RVVMF32BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 32) + (const_int 8)) + (eq_attr "mode" "RVVMF64BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 64) + (const_int 8)) (eq_attr "mode" "RVVM8HI,RVVM4HI,RVVM2HI,RVVM1HI,RVVMF2HI,RVVMF4HI,\ RVVM1x8HI,RVVMF2x8HI,RVVMF4x8HI,\ RVVM1x7HI,RVVMF2x7HI,RVVMF4x7HI,\ @@ -159,9 +171,9 @@ (eq_attr "mode" "RVVM4QI,RVVMF2BI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2QI,RVVMF4BI") (symbol_ref "riscv_vector::LMUL_2") (eq_attr "mode" "RVVM1QI,RVVMF8BI") (symbol_ref "riscv_vector::LMUL_1") - (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "riscv_vector::LMUL_F2") - (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "riscv_vector::LMUL_F4") - (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "riscv_vector::LMUL_F8") + (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F2") + (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F4") + (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F8") (eq_attr "mode" "RVVM8HI") (symbol_ref "riscv_vector::LMUL_8") (eq_attr "mode" "RVVM4HI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2HI") (symbol_ref "riscv_vector::LMUL_2") @@ -436,6 +448,10 @@ vaeskf1,vaeskf2,vaesz,vsha2ms,vsha2ch,vsha2cl,vsm4k,vsm4r,\ vsm3me,vsm3c") (const_int INVALID_ATTRIBUTE) + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE) (eq_attr "mode" "RVVM8QI,RVVM1BI") (const_int 1) (eq_attr "mode" "RVVM4QI,RVVMF2BI") (const_int 2) (eq_attr "mode" "RVVM2QI,RVVMF4BI") (const_int 4) @@ -905,6 +921,8 @@ (symbol_ref "riscv_vector::FRM_DYN")] (symbol_ref "riscv_vector::FRM_NONE"))) +(include "thead-vector.md") + ;; ----------------------------------------------------------------- ;; ---- Miscellaneous Operations ;; ----------------------------------------------------------------- @@ -1078,6 +1096,12 @@ - We can not leave it to TARGET_SECONDARY_RELOAD since it happens before spilling. The clobber scratch is used by spilling fractional registers in IRA/LRA so it's too early. */ + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; @@ -1114,7 +1138,7 @@ (define_insn "*mov<mode>_whole" [(set (match_operand:V_WHOLE 0 "reg_or_mem_operand" "=vr, m,vr") (match_operand:V_WHOLE 1 "reg_or_mem_operand" " m,vr,vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "@ vl%m1re<sew>.v\t%0,%1 vs%m1r.v\t%1,%0 @@ -1135,6 +1159,13 @@ (match_operand:VB 1 "general_operand"))] "TARGET_VECTOR" { + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } + if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; }) @@ -1142,7 +1173,7 @@ (define_insn "*mov<mode>" [(set (match_operand:VB 0 "register_operand" "=vr") (match_operand:VB 1 "register_operand" " vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "vmv1r.v\t%0,%1" [(set_attr "type" "vmov") (set_attr "mode" "<MODE>")]) @@ -3692,7 +3723,7 @@ (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3713,7 +3744,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3734,7 +3765,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c index 2e0e12aa045..2eef9e1e1a8 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c @@ -1,4 +1,4 @@ -/* { dg-do compile } */ +/* { dg-do compile { target { ! riscv_xtheadvector } } } */ /* { dg-skip-if "test rvv intrinsic" { *-*-* } { "*" } { "-march=rv*v*" } } */ void foo0 () {__rvv_bool64_t t;} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c index 3d81b179235..ef329e30785 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c @@ -1,4 +1,4 @@ /* { dg-do compile } */ /* { dg-options "-O3 -march=rv32gc -mabi=ilp32d" } */ -#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' extension enabled} } */ +#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' or 'XTHEADVECTOR' extension enabled} } */ diff --git a/gcc/testsuite/lib/target-supports.exp b/gcc/testsuite/lib/target-supports.exp index b27c30b8c51..bf0ad5342d6 100644 --- a/gcc/testsuite/lib/target-supports.exp +++ b/gcc/testsuite/lib/target-supports.exp @@ -1956,6 +1956,18 @@ proc check_effective_target_riscv_zbb { } { }] } +# Return 1 if the target arch supports the XTheadVector extension, 0 otherwise. +# Cache the result. + +proc check_effective_target_riscv_xtheadvector { } { + return [check_no_compiler_messages riscv_ext_xtheadvector assembly { + #ifndef __riscv_xtheadvector + #error "Not __riscv_xtheadvector" + #endif + }] +} + + # Return 1 if we can execute code when using dg-add-options riscv_v proc check_effective_target_riscv_v_ok { } { -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re：[PATCH v5] RISC-V: Handle differences between XTheadvector and Vector 2024-01-10 2:34 ` juzhe.zhong @ 2024-01-10 2:57 ` joshua 2024-01-10 3:02 ` juzhe.zhong ` (2 more replies) 0 siblings, 3 replies; 133+ messages in thread From: joshua @ 2024-01-10 2:57 UTC (permalink / raw) To: juzhe.zhong, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu Hi Juzhe, Thank you for so many useful comments for this patch! There are some more patches to support xtheadvector special instrinsics as well as handle register overlap issue and rewrite assembly output. https://gcc.gnu.org/pipermail/gcc-patches/2024-January/641774.html https://gcc.gnu.org/pipermail/gcc-patches/2024-January/641732.html https://gcc.gnu.org/pipermail/gcc-patches/2024-January/641733.html Also, there is a precedent patch to refactor riscv-vector-builtins-bases.cc https://gcc.gnu.org/pipermail/gcc-patches/2023-December/641530.html Jeff has reviewed it, but didn't have LGTM yet. Joshua ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月10日(星期三) 10:34 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Handle differences between XTheadvector and Vector Thanks for your patience. LGTM from myside. I think it's pretty clean now. I can image in the future when some day the theadvector is no longer used, we can remove it very easily. And also, the theadvector won't affect our RVV1.0 maintain since it's isolated cleanly. But I'd like to wait for a few more days some body want to chime in. And you should do more things before commit it: 1. Remember you should run the full coverage RVV1.0 API test, the test-generator is downloaded from official intrinsic doc: https://github.com/riscv-non-isa/rvv-intrinsic-doc 2. Also the regression of RV32 an RV64 of GCC testsuite. Do you have more patches of theadvector that I didn't review ? plz point them to me again. Thanks. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-10 10:22 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Handle differences between XTheadvector and Vector This patch is to handle the differences in instruction generation between Vector and XTheadVector. In this version, we only support partial xtheadvector instructions that leverage directly from current RVV1.0 with simple adding "th." prefix. For different name xtheadvector instructions but share same patterns as RVV1.0 instructions, we will use ASM targethook to rewrite the whole string of the instructions in the following patches. For some vector patterns that cannot be avoided, we use "!TARGET_XTHEADVECTOR" to disable them in vector.md in order not to generate instructions that xtheadvector does not support, like vmv1r and vsext.vf2. gcc/ChangeLog: * config.gcc: Add files for XTheadVector intrinsics. * config/riscv/autovec.md: Guard XTheadVector. * config/riscv/riscv-c.cc: Add pragma for XTheadVector. * config/riscv/riscv-string.cc (expand_block_move): Guard XTheadVector. * config/riscv/riscv-string.cc (vls_mode_valid_p): Avoid autovec. * config/riscv/riscv-vector-builtins-shapes.cc (check_type): (build_one): New function. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): (DEF_THEAD_RVV_FUNCTION): Add new marcos. (check_required_extensions): (handle_pragma_vector): * config/riscv/riscv-vector-builtins.h (RVV_REQUIRE_VECTOR): (RVV_REQUIRE_XTHEADVECTOR): Add RVV_REQUIRE_VECTOR and RVV_REQUIRE_XTHEADVECTOR. (struct function_group_info): * config/riscv/riscv-vector-switch.def (ENTRY): Disable fractional mode for the XTheadVector extension. (TUPLE_ENTRY): Likewise. * config/riscv/riscv-vsetvl.cc: Add functions for xtheadvector. * config/riscv/riscv.cc (riscv_v_ext_vls_mode_p): Guard XTheadVector. (riscv_v_adjust_bytesize): Likewise. (riscv_preferred_simd_mode): Likewsie. (riscv_autovectorize_vector_modes): Likewise. (riscv_vector_mode_supported_any_target_p): Likewise. (TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P): Likewise. * config/riscv/vector.md: Include thead-vector.md. * config/riscv/riscv_th_vector.h: New file. * config/riscv/thead-vector.md: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/base/pragma-1.c: Add XTheadVector. * gcc.target/riscv/rvv/base/abi-1.c: Exclude XTheadVector. * lib/target-supports.exp: Add target for XTheadVector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config.gcc | 2 +- gcc/config/riscv/autovec.md | 2 +- gcc/config/riscv/predicates.md | 4 +- gcc/config/riscv/riscv-c.cc | 3 +- gcc/config/riscv/riscv-string.cc | 3 +- gcc/config/riscv/riscv-v.cc | 2 +- .../riscv/riscv-vector-builtins-bases.cc | 48 ++++-- .../riscv/riscv-vector-builtins-shapes.cc | 23 +++ gcc/config/riscv/riscv-vector-switch.def | 150 +++++++++--------- gcc/config/riscv/riscv.cc | 20 ++- gcc/config/riscv/riscv_th_vector.h | 49 ++++++ gcc/config/riscv/thead-vector.md | 102 ++++++++++++ gcc/config/riscv/thead.cc | 23 ++- gcc/config/riscv/vector.md | 49 ++++-- .../gcc.target/riscv/rvv/base/abi-1.c | 2 +- .../gcc.target/riscv/rvv/base/pragma-1.c | 2 +- gcc/testsuite/lib/target-supports.exp | 12 ++ 17 files changed, 383 insertions(+), 113 deletions(-) create mode 100644 gcc/config/riscv/riscv_th_vector.h create mode 100644 gcc/config/riscv/thead-vector.md diff --git a/gcc/config.gcc b/gcc/config.gcc index 7e583390024..047e4c02cf4 100644 --- a/gcc/config.gcc +++ b/gcc/config.gcc @@ -549,7 +549,7 @@ riscv*) extra_objs="${extra_objs} riscv-vector-builtins.o riscv-vector-builtins-shapes.o riscv-vector-builtins-bases.o" extra_objs="${extra_objs} thead.o riscv-target-attr.o" d_target_objs="riscv-d.o" - extra_headers="riscv_vector.h" + extra_headers="riscv_vector.h riscv_th_vector.h" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.cc" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.h" ;; diff --git a/gcc/config/riscv/autovec.md b/gcc/config/riscv/autovec.md index 775eaa825b0..0477781cabe 100644 --- a/gcc/config/riscv/autovec.md +++ b/gcc/config/riscv/autovec.md @@ -2579,7 +2579,7 @@ [(match_operand 0 "register_operand") (match_operand 1 "memory_operand") (match_operand:ANYI 2 "const_int_operand")] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" { riscv_vector::expand_rawmemchr(<MODE>mode, operands[0], operands[1], operands[2]); diff --git a/gcc/config/riscv/predicates.md b/gcc/config/riscv/predicates.md index b1a79cae50a..0337da88284 100644 --- a/gcc/config/riscv/predicates.md +++ b/gcc/config/riscv/predicates.md @@ -428,7 +428,9 @@ ;; Predicates for the V extension. (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (ior (match_test "TARGET_XTHEADVECTOR && rtx_equal_p (op, const0_rtx)") + (match_test "!TARGET_XTHEADVECTOR")) + (match_operand 0 "const_csr_operand")))) (define_special_predicate "autovec_length_operand" (ior (match_operand 0 "pmode_register_operand") diff --git a/gcc/config/riscv/riscv-c.cc b/gcc/config/riscv/riscv-c.cc index 422ddc2c308..2e306057347 100644 --- a/gcc/config/riscv/riscv-c.cc +++ b/gcc/config/riscv/riscv-c.cc @@ -195,7 +195,8 @@ riscv_pragma_intrinsic (cpp_reader *) const char *name = TREE_STRING_POINTER (x); - if (strcmp (name, "vector") == 0) + if (strcmp (name, "vector") == 0 + || strcmp (name, "xtheadvector") == 0) { if (!TARGET_VECTOR) { diff --git a/gcc/config/riscv/riscv-string.cc b/gcc/config/riscv/riscv-string.cc index f227b8667ce..b09b51d7526 100644 --- a/gcc/config/riscv/riscv-string.cc +++ b/gcc/config/riscv/riscv-string.cc @@ -773,7 +773,8 @@ riscv_expand_block_move_scalar (rtx dest, rtx src, rtx length) bool riscv_expand_block_move (rtx dest, rtx src, rtx length) { - if (TARGET_VECTOR && stringop_strategy & STRATEGY_VECTOR) + if ((TARGET_VECTOR && !TARGET_XTHEADVECTOR) + && stringop_strategy & STRATEGY_VECTOR) { bool ok = riscv_vector::expand_block_move (dest, src, length); if (ok) diff --git a/gcc/config/riscv/riscv-v.cc b/gcc/config/riscv/riscv-v.cc index 2491522191a..7e25deb966d 100644 --- a/gcc/config/riscv/riscv-v.cc +++ b/gcc/config/riscv/riscv-v.cc @@ -4392,7 +4392,7 @@ cmp_lmul_gt_one (machine_mode mode) bool vls_mode_valid_p (machine_mode vls_mode) { - if (!TARGET_VECTOR) + if (!TARGET_VECTOR || TARGET_XTHEADVECTOR) return false; if (riscv_autovec_preference == RVV_SCALABLE) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 91af2c00674..46f1a1da33e 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -115,23 +115,37 @@ public: tree type = builtin_types[e.type.index].vector; machine_mode mode = TYPE_MODE (type); - /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. - - - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 - - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 - - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 - - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 - - e8,m2/e16,m4/e32,m8 --> e8m2 - - e8,m4/e16,m8 --> e8m4 - - e8,m8 --> e8m8 - */ - /* SEW. */ - e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); - - /* LMUL. */ - machine_mode e8_mode - = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); - e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + + if (TARGET_XTHEADVECTOR) + { + machine_mode inner_mode = GET_MODE_INNER (mode); + /* SEW. */ + e.add_input_operand (Pmode, + gen_int_mode (GET_MODE_BITSIZE (inner_mode), Pmode)); + /* LMUL. */ + e.add_input_operand (Pmode, + gen_int_mode (get_vlmul (mode), Pmode)); + } + else + { + /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. + + - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 + - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 + - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 + - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 + - e8,m2/e16,m4/e32,m8 --> e8m2 + - e8,m4/e16,m8 --> e8m4 + - e8,m8 --> e8m8 + */ + /* SEW. */ + e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); + + /* LMUL. */ + machine_mode e8_mode + = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); + e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + } /* TAIL_ANY. */ e.add_input_operand (Pmode, diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index ee8058dcdb1..1e4f4d53de6 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -33,6 +33,25 @@ namespace riscv_vector { +/* Check whether the RETURN_TYPE and ARGUMENT_TYPES are + valid for the function. */ + +static bool +check_type (tree return_type, vec<tree> &argument_types) +{ + tree arg; + unsigned i; + + if (!return_type) + return false; + + FOR_EACH_VEC_ELT (argument_types, i, arg) + if (!arg) + return false; + + return true; +} + /* Add one function instance for GROUP, using operand suffix at index OI, mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ static void @@ -49,6 +68,10 @@ build_one (function_builder &b, const function_group_info &group, group.ops_infos.types[vec_type_idx].index); b.allocate_argument_types (function_instance, argument_types); b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + b.add_overloaded_function (function_instance, *group.shape); b.add_unique_function (function_instance, (*group.shape), return_type, argument_types); diff --git a/gcc/config/riscv/riscv-vector-switch.def b/gcc/config/riscv/riscv-vector-switch.def index 1ad26c2a3b2..452283b7416 100644 --- a/gcc/config/riscv/riscv-vector-switch.def +++ b/gcc/config/riscv/riscv-vector-switch.def @@ -68,9 +68,9 @@ Encode the ratio of SEW/LMUL into the mask types. #endif /* Disable modes if TARGET_MIN_VLEN == 32. */ -ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) -ENTRY (RVVMF32BI, true, LMUL_F4, 32) -ENTRY (RVVMF16BI, true, LMUL_F2, 16) +ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F8, 64) +ENTRY (RVVMF32BI, true, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F4, 32) +ENTRY (RVVMF16BI, true, TARGET_XTHEADVECTOR ? LMUL_1 : LMUL_F2 , 16) ENTRY (RVVMF8BI, true, LMUL_1, 8) ENTRY (RVVMF4BI, true, LMUL_2, 4) ENTRY (RVVMF2BI, true, LMUL_4, 2) @@ -81,39 +81,39 @@ ENTRY (RVVM8QI, true, LMUL_8, 1) ENTRY (RVVM4QI, true, LMUL_4, 2) ENTRY (RVVM2QI, true, LMUL_2, 4) ENTRY (RVVM1QI, true, LMUL_1, 8) -ENTRY (RVVMF2QI, true, LMUL_F2, 16) -ENTRY (RVVMF4QI, true, LMUL_F4, 32) -ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) +ENTRY (RVVMF2QI, !TARGET_XTHEADVECTOR, LMUL_F2, 16) +ENTRY (RVVMF4QI, !TARGET_XTHEADVECTOR, LMUL_F4, 32) +ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F8, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8HI, true, LMUL_8, 2) ENTRY (RVVM4HI, true, LMUL_4, 4) ENTRY (RVVM2HI, true, LMUL_2, 8) ENTRY (RVVM1HI, true, LMUL_1, 16) -ENTRY (RVVMF2HI, true, LMUL_F2, 32) -ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HI, !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_16. */ ENTRY (RVVM8HF, TARGET_VECTOR_ELEN_FP_16, LMUL_8, 2) ENTRY (RVVM4HF, TARGET_VECTOR_ELEN_FP_16, LMUL_4, 4) ENTRY (RVVM2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_2, 8) ENTRY (RVVM1HF, TARGET_VECTOR_ELEN_FP_16, LMUL_1, 16) -ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_F2, 32) -ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8SI, true, LMUL_8, 4) ENTRY (RVVM4SI, true, LMUL_4, 8) ENTRY (RVVM2SI, true, LMUL_2, 16) ENTRY (RVVM1SI, true, LMUL_1, 32) -ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_32. */ ENTRY (RVVM8SF, TARGET_VECTOR_ELEN_FP_32, LMUL_8, 4) ENTRY (RVVM4SF, TARGET_VECTOR_ELEN_FP_32, LMUL_4, 8) ENTRY (RVVM2SF, TARGET_VECTOR_ELEN_FP_32, LMUL_2, 16) ENTRY (RVVM1SF, TARGET_VECTOR_ELEN_FP_32, LMUL_1, 32) -ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if !TARGET_VECTOR_ELEN_64. */ ENTRY (RVVM8DI, TARGET_VECTOR_ELEN_64, LMUL_8, 8) @@ -140,127 +140,127 @@ ENTRY (RVVM1DF, TARGET_VECTOR_ELEN_FP_64, LMUL_1, 64) #endif TUPLE_ENTRY (RVVM1x8QI, true, RVVM1QI, 8, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x8QI, true, RVVMF2QI, 8, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x8QI, true, RVVMF4QI, 8, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 8, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x8QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 8, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x8QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 8, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 8, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x7QI, true, RVVM1QI, 7, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x7QI, true, RVVMF2QI, 7, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x7QI, true, RVVMF4QI, 7, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 7, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x7QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 7, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x7QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 7, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 7, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x6QI, true, RVVM1QI, 6, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x6QI, true, RVVMF2QI, 6, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x6QI, true, RVVMF4QI, 6, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 6, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x6QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 6, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x6QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 6, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 6, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x5QI, true, RVVM1QI, 5, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x5QI, true, RVVMF2QI, 5, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x5QI, true, RVVMF4QI, 5, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 5, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x5QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 5, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x5QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 5, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 5, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x4QI, true, RVVM2QI, 4, LMUL_2, 4) TUPLE_ENTRY (RVVM1x4QI, true, RVVM1QI, 4, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x4QI, true, RVVMF2QI, 4, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x4QI, true, RVVMF4QI, 4, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 4, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x4QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 4, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x4QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 4, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 4, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x3QI, true, RVVM2QI, 3, LMUL_2, 4) TUPLE_ENTRY (RVVM1x3QI, true, RVVM1QI, 3, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x3QI, true, RVVMF2QI, 3, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x3QI, true, RVVMF4QI, 3, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 3, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x3QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 3, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x3QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 3, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 3, LMUL_F8, 64) TUPLE_ENTRY (RVVM4x2QI, true, RVVM4QI, 2, LMUL_4, 2) TUPLE_ENTRY (RVVM2x2QI, true, RVVM2QI, 2, LMUL_2, 4) TUPLE_ENTRY (RVVM1x2QI, true, RVVM1QI, 2, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x2QI, true, RVVMF2QI, 2, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x2QI, true, RVVMF4QI, 2, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 2, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x2QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 2, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x2QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 2, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 2, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x8HI, true, RVVM1HI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HI, true, RVVMF2HI, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HI, true, RVVM1HI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HI, true, RVVMF2HI, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HI, true, RVVM1HI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HI, true, RVVMF2HI, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HI, true, RVVM1HI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HI, true, RVVMF2HI, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HI, true, RVVM2HI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HI, true, RVVM1HI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HI, true, RVVMF2HI, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HI, true, RVVM2HI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HI, true, RVVM1HI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HI, true, RVVMF2HI, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HI, true, RVVM4HI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HI, true, RVVM2HI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HI, true, RVVM1HI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HI, true, RVVMF2HI, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM4HF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8SI, true, RVVM1SI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SI, true, RVVM1SI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SI, true, RVVM1SI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SI, true, RVVM1SI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SI, true, RVVM2SI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SI, true, RVVM1SI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SI, true, RVVM2SI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SI, true, RVVM1SI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SI, true, RVVM4SI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SI, true, RVVM2SI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SI, true, RVVM1SI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM4SF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 8, LMUL_1, 16) TUPLE_ENTRY (RVVM1x7DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 7, LMUL_1, 16) diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index 8f74cd2e502..28cf1f414e9 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -1406,6 +1406,9 @@ riscv_v_adjust_bytesize (machine_mode mode, int scale) { if (riscv_v_ext_vector_mode_p (mode)) { + if (TARGET_XTHEADVECTOR) + return BYTES_PER_RISCV_VECTOR; + poly_int64 nunits = GET_MODE_NUNITS (mode); poly_int64 mode_size = GET_MODE_SIZE (mode); @@ -9995,7 +9998,7 @@ riscv_use_divmod_expander (void) static machine_mode riscv_preferred_simd_mode (scalar_mode mode) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::preferred_simd_mode (mode); return word_mode; @@ -10346,7 +10349,7 @@ riscv_mode_priority (int, int n) unsigned int riscv_autovectorize_vector_modes (vector_modes *modes, bool all) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::autovectorize_vector_modes (modes, all); return default_autovectorize_vector_modes (modes, all); @@ -10529,6 +10532,16 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) return false; } +/* Implements target hook vector_mode_supported_any_target_p. */ + +static bool +riscv_vector_mode_supported_any_target_p (machine_mode mode) +{ + if (TARGET_XTHEADVECTOR) + return false; + return true; +} + /* Initialize the GCC target structure. */ #undef TARGET_ASM_ALIGNED_HI_OP #define TARGET_ASM_ALIGNED_HI_OP "\t.half\t" @@ -10872,6 +10885,9 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) #undef TARGET_PREFERRED_ELSE_VALUE #define TARGET_PREFERRED_ELSE_VALUE riscv_preferred_else_value +#undef TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P +#define TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P riscv_vector_mode_supported_any_target_p + struct gcc_target targetm = TARGET_INITIALIZER; #include "gt-riscv.h" diff --git a/gcc/config/riscv/riscv_th_vector.h b/gcc/config/riscv/riscv_th_vector.h new file mode 100644 index 00000000000..b6b6738bdda --- /dev/null +++ b/gcc/config/riscv/riscv_th_vector.h @@ -0,0 +1,49 @@ +/* RISC-V 'XTheadVector' Extension intrinsics include file. + Copyright (C) 2024 Free Software Foundation, Inc. + + This file is part of GCC. + + GCC is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published + by the Free Software Foundation; either version 3, or (at your + option) any later version. + + GCC is distributed in the hope that it will be useful, but WITHOUT + ANY WARRANTY; without even the implied warranty of MERCHANTABILITY + or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public + License for more details. + + Under Section 7 of GPL version 3, you are granted additional + permissions described in the GCC Runtime Library Exception, version + 3.1, as published by the Free Software Foundation. + + You should have received a copy of the GNU General Public License and + a copy of the GCC Runtime Library Exception along with this program; + see the files COPYING3 and COPYING.RUNTIME respectively. If not, see + <" target="_blank">http://www.gnu.org/licenses/>. */ + +#ifndef __RISCV_TH_VECTOR_H +#define __RISCV_TH_VECTOR_H + +#include <stdint.h> +#include <stddef.h> + +#ifndef __riscv_xtheadvector +#error "XTheadVector intrinsics require the xtheadvector extension." +#else +#ifdef __cplusplus +extern "C" { +#endif + +/* NOTE: This implementation of riscv_th_vector.h is intentionally short. It does + not define the RVV types and intrinsic functions directly in C and C++ + code, but instead uses the following pragma to tell GCC to insert the + necessary type and function definitions itself. The net effect is the + same, and the file is a complete implementation of riscv_th_vector.h. */ +#pragma riscv intrinsic "xtheadvector" + +#ifdef __cplusplus +} +#endif // __cplusplus +#endif // __riscv_xtheadvector +#endif // __RISCV_TH_ECTOR_H diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md new file mode 100644 index 00000000000..696b815252d --- /dev/null +++ b/gcc/config/riscv/thead-vector.md @@ -0,0 +1,102 @@ +(define_c_enum "unspec" [ + UNSPEC_TH_VWLDST +]) + +(define_mode_iterator V_VLS_VT [V VLS VT]) +(define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) + +(define_split + [(set (match_operand:V_VB_VLS_VT 0 "reg_or_mem_operand") + (match_operand:V_VB_VLS_VT 1 "reg_or_mem_operand"))] + "TARGET_XTHEADVECTOR" + [(const_int 0)] + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + }) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:V_VLS_VT 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:V_VLS_VT + [(match_operand:V_VLS_VT 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2")]) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:VB 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:VB + [(match_operand:VB 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2") + (set (attr "sew") (const_int 8)) + (set (attr "vlmul") (symbol_ref "riscv_vector::LMUL_1"))]) + +(define_insn_and_split "*pred_th_mov<mode>" + [(set (match_operand:VB_VLS 0 "nonimmediate_operand" "=vr, m, vr, vr, vr") + (if_then_else:VB_VLS + (unspec:VB_VLS + [(match_operand:VB_VLS 1 "vector_all_trues_mask_operand" "Wc1, Wc1, Wc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) + (match_operand:VB_VLS 3 "vector_move_operand" " m, vr, vr, Wc0, Wc1") + (match_operand:VB_VLS 2 "vector_undef_operand" " vu, vu, vu, vu, vu")))] + "TARGET_XTHEADVECTOR" + "@ + # + # + vmcpy.m\t%0,%3 + vmclr.m\t%0 + vmset.m\t%0" + "&& !reload_completed" + [(const_int 0)] + { + if ((MEM_P (operands[0]) || MEM_P (operands[3])) + || (REG_P (operands[0]) && REG_P (operands[3]) + && INTVAL (operands[5]) == riscv_vector::VLMAX)) + { + emit_move_insn (operands[0], operands[3]); + DONE; + } + FAIL; + } + [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/config/riscv/thead.cc b/gcc/config/riscv/thead.cc index 00b2e14c6c0..cc04101b256 100644 --- a/gcc/config/riscv/thead.cc +++ b/gcc/config/riscv/thead.cc @@ -27,6 +27,7 @@ #include "backend.h" #include "tree.h" #include "rtl.h" +#include "insn-attr.h" #include "explow.h" #include "memmodel.h" #include "emit-rtl.h" @@ -890,8 +891,26 @@ th_asm_output_opcode (FILE *asm_out_file, const char *p) { /* We need to add th. prefix to all the xtheadvector instructions here.*/ - if (current_output_insn != NULL && p[0] == 'v') - fputs ("th.", asm_out_file); + if (current_output_insn != NULL) + { + if (get_attr_type (current_output_insn) == TYPE_VSETVL) + { + if (strstr (p, "zero")) + { + if (strstr (p, "zero,zero")) + return "th.vsetvli\tzero,zero,e%0,%m1"; + else + return "th.vsetvli\tzero,%0,e%1,%m2"; + } + else + { + return "th.vsetvli\t%0,%1,e%2,%m3"; + } + } + + if (p[0] == 'v') + fputs ("th.", asm_out_file); + } return p; } diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index 24b7b4394be..3eb6daafbc2 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -87,7 +87,7 @@ ;; check. However, we need default value of SEW for vsetvl instruction since there ;; is no field for ratio in the vsetvl instruction encoding. (define_attr "sew" "" - (cond [(eq_attr "mode" "RVVMF64BI,RVVMF32BI,RVVMF16BI,RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ + (cond [(eq_attr "mode" "RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ RVVM8QI,RVVM4QI,RVVM2QI,RVVM1QI,RVVMF2QI,RVVMF4QI,RVVMF8QI,\ RVVM1x8QI,RVVMF2x8QI,RVVMF4x8QI,RVVMF8x8QI,\ RVVM1x7QI,RVVMF2x7QI,RVVMF4x7QI,RVVMF8x7QI,\ @@ -99,6 +99,18 @@ V1QI,V2QI,V4QI,V8QI,V16QI,V32QI,V64QI,V128QI,V256QI,V512QI,V1024QI,V2048QI,V4096QI,\ V1BI,V2BI,V4BI,V8BI,V16BI,V32BI,V64BI,V128BI,V256BI,V512BI,V1024BI,V2048BI,V4096BI") (const_int 8) + (eq_attr "mode" "RVVMF16BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 16) + (const_int 8)) + (eq_attr "mode" "RVVMF32BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 32) + (const_int 8)) + (eq_attr "mode" "RVVMF64BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 64) + (const_int 8)) (eq_attr "mode" "RVVM8HI,RVVM4HI,RVVM2HI,RVVM1HI,RVVMF2HI,RVVMF4HI,\ RVVM1x8HI,RVVMF2x8HI,RVVMF4x8HI,\ RVVM1x7HI,RVVMF2x7HI,RVVMF4x7HI,\ @@ -159,9 +171,9 @@ (eq_attr "mode" "RVVM4QI,RVVMF2BI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2QI,RVVMF4BI") (symbol_ref "riscv_vector::LMUL_2") (eq_attr "mode" "RVVM1QI,RVVMF8BI") (symbol_ref "riscv_vector::LMUL_1") - (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "riscv_vector::LMUL_F2") - (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "riscv_vector::LMUL_F4") - (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "riscv_vector::LMUL_F8") + (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F2") + (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F4") + (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F8") (eq_attr "mode" "RVVM8HI") (symbol_ref "riscv_vector::LMUL_8") (eq_attr "mode" "RVVM4HI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2HI") (symbol_ref "riscv_vector::LMUL_2") @@ -436,6 +448,10 @@ vaeskf1,vaeskf2,vaesz,vsha2ms,vsha2ch,vsha2cl,vsm4k,vsm4r,\ vsm3me,vsm3c") (const_int INVALID_ATTRIBUTE) + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE) (eq_attr "mode" "RVVM8QI,RVVM1BI") (const_int 1) (eq_attr "mode" "RVVM4QI,RVVMF2BI") (const_int 2) (eq_attr "mode" "RVVM2QI,RVVMF4BI") (const_int 4) @@ -905,6 +921,8 @@ (symbol_ref "riscv_vector::FRM_DYN")] (symbol_ref "riscv_vector::FRM_NONE"))) +(include "thead-vector.md") + ;; ----------------------------------------------------------------- ;; ---- Miscellaneous Operations ;; ----------------------------------------------------------------- @@ -1078,6 +1096,12 @@ - We can not leave it to TARGET_SECONDARY_RELOAD since it happens before spilling. The clobber scratch is used by spilling fractional registers in IRA/LRA so it's too early. */ + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; @@ -1114,7 +1138,7 @@ (define_insn "*mov<mode>_whole" [(set (match_operand:V_WHOLE 0 "reg_or_mem_operand" "=vr, m,vr") (match_operand:V_WHOLE 1 "reg_or_mem_operand" " m,vr,vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "@ vl%m1re<sew>.v\t%0,%1 vs%m1r.v\t%1,%0 @@ -1135,6 +1159,13 @@ (match_operand:VB 1 "general_operand"))] "TARGET_VECTOR" { + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } + if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; }) @@ -1142,7 +1173,7 @@ (define_insn "*mov<mode>" [(set (match_operand:VB 0 "register_operand" "=vr") (match_operand:VB 1 "register_operand" " vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "vmv1r.v\t%0,%1" [(set_attr "type" "vmov") (set_attr "mode" "<MODE>")]) @@ -3692,7 +3723,7 @@ (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3713,7 +3744,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3734,7 +3765,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c index 2e0e12aa045..2eef9e1e1a8 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c @@ -1,4 +1,4 @@ -/* { dg-do compile } */ +/* { dg-do compile { target { ! riscv_xtheadvector } } } */ /* { dg-skip-if "test rvv intrinsic" { *-*-* } { "*" } { "-march=rv*v*" } } */ void foo0 () {__rvv_bool64_t t;} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c index 3d81b179235..ef329e30785 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c @@ -1,4 +1,4 @@ /* { dg-do compile } */ /* { dg-options "-O3 -march=rv32gc -mabi=ilp32d" } */ -#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' extension enabled} } */ +#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' or 'XTHEADVECTOR' extension enabled} } */ diff --git a/gcc/testsuite/lib/target-supports.exp b/gcc/testsuite/lib/target-supports.exp index b27c30b8c51..bf0ad5342d6 100644 --- a/gcc/testsuite/lib/target-supports.exp +++ b/gcc/testsuite/lib/target-supports.exp @@ -1956,6 +1956,18 @@ proc check_effective_target_riscv_zbb { } { }] } +# Return 1 if the target arch supports the XTheadVector extension, 0 otherwise. +# Cache the result. + +proc check_effective_target_riscv_xtheadvector { } { + return [check_no_compiler_messages riscv_ext_xtheadvector assembly { + #ifndef __riscv_xtheadvector + #error "Not __riscv_xtheadvector" + #endif + }] +} + + # Return 1 if we can execute code when using dg-add-options riscv_v proc check_effective_target_riscv_v_ok { } { -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: Re：[PATCH v5] RISC-V: Handle differences between XTheadvector and Vector 2024-01-10 2:57 ` Re：[PATCH " joshua @ 2024-01-10 3:02 ` juzhe.zhong 2024-01-10 6:54 ` juzhe.zhong 2024-01-10 7:01 ` juzhe.zhong 2 siblings, 0 replies; 133+ messages in thread From: juzhe.zhong @ 2024-01-10 3:02 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 50175 bytes --] ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr, &vr, &vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") - (match_operand 7 "const_int_operand" " i, i, i, i") - (match_operand 8 "const_int_operand" " i, i, i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK, rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i, i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_ltge_operator" - [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr") - (match_operand:V_VLSI 5 "vector_arith_operand" " vr, vr, vi, vi")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] + [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr, vr, vr, vr, vr") + (match_operand:V_VLSI 5 "vector_arith_operand" " vr, vr, vi, vi, vr, vr, vi, vi")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0, vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,th,th,none,none,none,none")]) You are add "&vr, &vr, &vr, &vr" which will be enabled when TARGET_VECTOR. You should disable these constraints when TARGET_VECTOR is enabled. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-10 10:57 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Handle differences between XTheadvector and Vector Hi Juzhe, Thank you for so many useful comments for this patch! There are some more patches to support xtheadvector special instrinsics as well as handle register overlap issue and rewrite assembly output. https://gcc.gnu.org/pipermail/gcc-patches/2024-January/641774.html https://gcc.gnu.org/pipermail/gcc-patches/2024-January/641732.html https://gcc.gnu.org/pipermail/gcc-patches/2024-January/641733.html Also, there is a precedent patch to refactor riscv-vector-builtins-bases.cc https://gcc.gnu.org/pipermail/gcc-patches/2023-December/641530.html Jeff has reviewed it, but didn't have LGTM yet. Joshua ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月10日(星期三) 10:34 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Handle differences between XTheadvector and Vector Thanks for your patience. LGTM from myside. I think it's pretty clean now. I can image in the future when some day the theadvector is no longer used, we can remove it very easily. And also, the theadvector won't affect our RVV1.0 maintain since it's isolated cleanly. But I'd like to wait for a few more days some body want to chime in. And you should do more things before commit it: 1. Remember you should run the full coverage RVV1.0 API test, the test-generator is downloaded from official intrinsic doc: https://github.com/riscv-non-isa/rvv-intrinsic-doc 2. Also the regression of RV32 an RV64 of GCC testsuite. Do you have more patches of theadvector that I didn't review ? plz point them to me again. Thanks. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-10 10:22 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Handle differences between XTheadvector and Vector This patch is to handle the differences in instruction generation between Vector and XTheadVector. In this version, we only support partial xtheadvector instructions that leverage directly from current RVV1.0 with simple adding "th." prefix. For different name xtheadvector instructions but share same patterns as RVV1.0 instructions, we will use ASM targethook to rewrite the whole string of the instructions in the following patches. For some vector patterns that cannot be avoided, we use "!TARGET_XTHEADVECTOR" to disable them in vector.md in order not to generate instructions that xtheadvector does not support, like vmv1r and vsext.vf2. gcc/ChangeLog: * config.gcc: Add files for XTheadVector intrinsics. * config/riscv/autovec.md: Guard XTheadVector. * config/riscv/riscv-c.cc: Add pragma for XTheadVector. * config/riscv/riscv-string.cc (expand_block_move): Guard XTheadVector. * config/riscv/riscv-string.cc (vls_mode_valid_p): Avoid autovec. * config/riscv/riscv-vector-builtins-shapes.cc (check_type): (build_one): New function. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): (DEF_THEAD_RVV_FUNCTION): Add new marcos. (check_required_extensions): (handle_pragma_vector): * config/riscv/riscv-vector-builtins.h (RVV_REQUIRE_VECTOR): (RVV_REQUIRE_XTHEADVECTOR): Add RVV_REQUIRE_VECTOR and RVV_REQUIRE_XTHEADVECTOR. (struct function_group_info): * config/riscv/riscv-vector-switch.def (ENTRY): Disable fractional mode for the XTheadVector extension. (TUPLE_ENTRY): Likewise. * config/riscv/riscv-vsetvl.cc: Add functions for xtheadvector. * config/riscv/riscv.cc (riscv_v_ext_vls_mode_p): Guard XTheadVector. (riscv_v_adjust_bytesize): Likewise. (riscv_preferred_simd_mode): Likewsie. (riscv_autovectorize_vector_modes): Likewise. (riscv_vector_mode_supported_any_target_p): Likewise. (TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P): Likewise. * config/riscv/vector.md: Include thead-vector.md. * config/riscv/riscv_th_vector.h: New file. * config/riscv/thead-vector.md: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/base/pragma-1.c: Add XTheadVector. * gcc.target/riscv/rvv/base/abi-1.c: Exclude XTheadVector. * lib/target-supports.exp: Add target for XTheadVector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config.gcc | 2 +- gcc/config/riscv/autovec.md | 2 +- gcc/config/riscv/predicates.md | 4 +- gcc/config/riscv/riscv-c.cc | 3 +- gcc/config/riscv/riscv-string.cc | 3 +- gcc/config/riscv/riscv-v.cc | 2 +- .../riscv/riscv-vector-builtins-bases.cc | 48 ++++-- .../riscv/riscv-vector-builtins-shapes.cc | 23 +++ gcc/config/riscv/riscv-vector-switch.def | 150 +++++++++--------- gcc/config/riscv/riscv.cc | 20 ++- gcc/config/riscv/riscv_th_vector.h | 49 ++++++ gcc/config/riscv/thead-vector.md | 102 ++++++++++++ gcc/config/riscv/thead.cc | 23 ++- gcc/config/riscv/vector.md | 49 ++++-- .../gcc.target/riscv/rvv/base/abi-1.c | 2 +- .../gcc.target/riscv/rvv/base/pragma-1.c | 2 +- gcc/testsuite/lib/target-supports.exp | 12 ++ 17 files changed, 383 insertions(+), 113 deletions(-) create mode 100644 gcc/config/riscv/riscv_th_vector.h create mode 100644 gcc/config/riscv/thead-vector.md diff --git a/gcc/config.gcc b/gcc/config.gcc index 7e583390024..047e4c02cf4 100644 --- a/gcc/config.gcc +++ b/gcc/config.gcc @@ -549,7 +549,7 @@ riscv*) extra_objs="${extra_objs} riscv-vector-builtins.o riscv-vector-builtins-shapes.o riscv-vector-builtins-bases.o" extra_objs="${extra_objs} thead.o riscv-target-attr.o" d_target_objs="riscv-d.o" - extra_headers="riscv_vector.h" + extra_headers="riscv_vector.h riscv_th_vector.h" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.cc" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.h" ;; diff --git a/gcc/config/riscv/autovec.md b/gcc/config/riscv/autovec.md index 775eaa825b0..0477781cabe 100644 --- a/gcc/config/riscv/autovec.md +++ b/gcc/config/riscv/autovec.md @@ -2579,7 +2579,7 @@ [(match_operand 0 "register_operand") (match_operand 1 "memory_operand") (match_operand:ANYI 2 "const_int_operand")] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" { riscv_vector::expand_rawmemchr(<MODE>mode, operands[0], operands[1], operands[2]); diff --git a/gcc/config/riscv/predicates.md b/gcc/config/riscv/predicates.md index b1a79cae50a..0337da88284 100644 --- a/gcc/config/riscv/predicates.md +++ b/gcc/config/riscv/predicates.md @@ -428,7 +428,9 @@ ;; Predicates for the V extension. (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (ior (match_test "TARGET_XTHEADVECTOR && rtx_equal_p (op, const0_rtx)") + (match_test "!TARGET_XTHEADVECTOR")) + (match_operand 0 "const_csr_operand")))) (define_special_predicate "autovec_length_operand" (ior (match_operand 0 "pmode_register_operand") diff --git a/gcc/config/riscv/riscv-c.cc b/gcc/config/riscv/riscv-c.cc index 422ddc2c308..2e306057347 100644 --- a/gcc/config/riscv/riscv-c.cc +++ b/gcc/config/riscv/riscv-c.cc @@ -195,7 +195,8 @@ riscv_pragma_intrinsic (cpp_reader *) const char *name = TREE_STRING_POINTER (x); - if (strcmp (name, "vector") == 0) + if (strcmp (name, "vector") == 0 + || strcmp (name, "xtheadvector") == 0) { if (!TARGET_VECTOR) { diff --git a/gcc/config/riscv/riscv-string.cc b/gcc/config/riscv/riscv-string.cc index f227b8667ce..b09b51d7526 100644 --- a/gcc/config/riscv/riscv-string.cc +++ b/gcc/config/riscv/riscv-string.cc @@ -773,7 +773,8 @@ riscv_expand_block_move_scalar (rtx dest, rtx src, rtx length) bool riscv_expand_block_move (rtx dest, rtx src, rtx length) { - if (TARGET_VECTOR && stringop_strategy & STRATEGY_VECTOR) + if ((TARGET_VECTOR && !TARGET_XTHEADVECTOR) + && stringop_strategy & STRATEGY_VECTOR) { bool ok = riscv_vector::expand_block_move (dest, src, length); if (ok) diff --git a/gcc/config/riscv/riscv-v.cc b/gcc/config/riscv/riscv-v.cc index 2491522191a..7e25deb966d 100644 --- a/gcc/config/riscv/riscv-v.cc +++ b/gcc/config/riscv/riscv-v.cc @@ -4392,7 +4392,7 @@ cmp_lmul_gt_one (machine_mode mode) bool vls_mode_valid_p (machine_mode vls_mode) { - if (!TARGET_VECTOR) + if (!TARGET_VECTOR || TARGET_XTHEADVECTOR) return false; if (riscv_autovec_preference == RVV_SCALABLE) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 91af2c00674..46f1a1da33e 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -115,23 +115,37 @@ public: tree type = builtin_types[e.type.index].vector; machine_mode mode = TYPE_MODE (type); - /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. - - - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 - - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 - - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 - - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 - - e8,m2/e16,m4/e32,m8 --> e8m2 - - e8,m4/e16,m8 --> e8m4 - - e8,m8 --> e8m8 - */ - /* SEW. */ - e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); - - /* LMUL. */ - machine_mode e8_mode - = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); - e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + + if (TARGET_XTHEADVECTOR) + { + machine_mode inner_mode = GET_MODE_INNER (mode); + /* SEW. */ + e.add_input_operand (Pmode, + gen_int_mode (GET_MODE_BITSIZE (inner_mode), Pmode)); + /* LMUL. */ + e.add_input_operand (Pmode, + gen_int_mode (get_vlmul (mode), Pmode)); + } + else + { + /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. + + - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 + - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 + - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 + - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 + - e8,m2/e16,m4/e32,m8 --> e8m2 + - e8,m4/e16,m8 --> e8m4 + - e8,m8 --> e8m8 + */ + /* SEW. */ + e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); + + /* LMUL. */ + machine_mode e8_mode + = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); + e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + } /* TAIL_ANY. */ e.add_input_operand (Pmode, diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index ee8058dcdb1..1e4f4d53de6 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -33,6 +33,25 @@ namespace riscv_vector { +/* Check whether the RETURN_TYPE and ARGUMENT_TYPES are + valid for the function. */ + +static bool +check_type (tree return_type, vec<tree> &argument_types) +{ + tree arg; + unsigned i; + + if (!return_type) + return false; + + FOR_EACH_VEC_ELT (argument_types, i, arg) + if (!arg) + return false; + + return true; +} + /* Add one function instance for GROUP, using operand suffix at index OI, mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ static void @@ -49,6 +68,10 @@ build_one (function_builder &b, const function_group_info &group, group.ops_infos.types[vec_type_idx].index); b.allocate_argument_types (function_instance, argument_types); b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + b.add_overloaded_function (function_instance, *group.shape); b.add_unique_function (function_instance, (*group.shape), return_type, argument_types); diff --git a/gcc/config/riscv/riscv-vector-switch.def b/gcc/config/riscv/riscv-vector-switch.def index 1ad26c2a3b2..452283b7416 100644 --- a/gcc/config/riscv/riscv-vector-switch.def +++ b/gcc/config/riscv/riscv-vector-switch.def @@ -68,9 +68,9 @@ Encode the ratio of SEW/LMUL into the mask types. #endif /* Disable modes if TARGET_MIN_VLEN == 32. */ -ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) -ENTRY (RVVMF32BI, true, LMUL_F4, 32) -ENTRY (RVVMF16BI, true, LMUL_F2, 16) +ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F8, 64) +ENTRY (RVVMF32BI, true, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F4, 32) +ENTRY (RVVMF16BI, true, TARGET_XTHEADVECTOR ? LMUL_1 : LMUL_F2 , 16) ENTRY (RVVMF8BI, true, LMUL_1, 8) ENTRY (RVVMF4BI, true, LMUL_2, 4) ENTRY (RVVMF2BI, true, LMUL_4, 2) @@ -81,39 +81,39 @@ ENTRY (RVVM8QI, true, LMUL_8, 1) ENTRY (RVVM4QI, true, LMUL_4, 2) ENTRY (RVVM2QI, true, LMUL_2, 4) ENTRY (RVVM1QI, true, LMUL_1, 8) -ENTRY (RVVMF2QI, true, LMUL_F2, 16) -ENTRY (RVVMF4QI, true, LMUL_F4, 32) -ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) +ENTRY (RVVMF2QI, !TARGET_XTHEADVECTOR, LMUL_F2, 16) +ENTRY (RVVMF4QI, !TARGET_XTHEADVECTOR, LMUL_F4, 32) +ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F8, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8HI, true, LMUL_8, 2) ENTRY (RVVM4HI, true, LMUL_4, 4) ENTRY (RVVM2HI, true, LMUL_2, 8) ENTRY (RVVM1HI, true, LMUL_1, 16) -ENTRY (RVVMF2HI, true, LMUL_F2, 32) -ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HI, !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_16. */ ENTRY (RVVM8HF, TARGET_VECTOR_ELEN_FP_16, LMUL_8, 2) ENTRY (RVVM4HF, TARGET_VECTOR_ELEN_FP_16, LMUL_4, 4) ENTRY (RVVM2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_2, 8) ENTRY (RVVM1HF, TARGET_VECTOR_ELEN_FP_16, LMUL_1, 16) -ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_F2, 32) -ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8SI, true, LMUL_8, 4) ENTRY (RVVM4SI, true, LMUL_4, 8) ENTRY (RVVM2SI, true, LMUL_2, 16) ENTRY (RVVM1SI, true, LMUL_1, 32) -ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_32. */ ENTRY (RVVM8SF, TARGET_VECTOR_ELEN_FP_32, LMUL_8, 4) ENTRY (RVVM4SF, TARGET_VECTOR_ELEN_FP_32, LMUL_4, 8) ENTRY (RVVM2SF, TARGET_VECTOR_ELEN_FP_32, LMUL_2, 16) ENTRY (RVVM1SF, TARGET_VECTOR_ELEN_FP_32, LMUL_1, 32) -ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if !TARGET_VECTOR_ELEN_64. */ ENTRY (RVVM8DI, TARGET_VECTOR_ELEN_64, LMUL_8, 8) @@ -140,127 +140,127 @@ ENTRY (RVVM1DF, TARGET_VECTOR_ELEN_FP_64, LMUL_1, 64) #endif TUPLE_ENTRY (RVVM1x8QI, true, RVVM1QI, 8, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x8QI, true, RVVMF2QI, 8, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x8QI, true, RVVMF4QI, 8, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 8, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x8QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 8, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x8QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 8, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 8, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x7QI, true, RVVM1QI, 7, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x7QI, true, RVVMF2QI, 7, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x7QI, true, RVVMF4QI, 7, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 7, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x7QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 7, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x7QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 7, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 7, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x6QI, true, RVVM1QI, 6, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x6QI, true, RVVMF2QI, 6, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x6QI, true, RVVMF4QI, 6, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 6, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x6QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 6, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x6QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 6, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 6, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x5QI, true, RVVM1QI, 5, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x5QI, true, RVVMF2QI, 5, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x5QI, true, RVVMF4QI, 5, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 5, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x5QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 5, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x5QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 5, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 5, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x4QI, true, RVVM2QI, 4, LMUL_2, 4) TUPLE_ENTRY (RVVM1x4QI, true, RVVM1QI, 4, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x4QI, true, RVVMF2QI, 4, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x4QI, true, RVVMF4QI, 4, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 4, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x4QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 4, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x4QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 4, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 4, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x3QI, true, RVVM2QI, 3, LMUL_2, 4) TUPLE_ENTRY (RVVM1x3QI, true, RVVM1QI, 3, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x3QI, true, RVVMF2QI, 3, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x3QI, true, RVVMF4QI, 3, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 3, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x3QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 3, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x3QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 3, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 3, LMUL_F8, 64) TUPLE_ENTRY (RVVM4x2QI, true, RVVM4QI, 2, LMUL_4, 2) TUPLE_ENTRY (RVVM2x2QI, true, RVVM2QI, 2, LMUL_2, 4) TUPLE_ENTRY (RVVM1x2QI, true, RVVM1QI, 2, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x2QI, true, RVVMF2QI, 2, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x2QI, true, RVVMF4QI, 2, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 2, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x2QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 2, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x2QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 2, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 2, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x8HI, true, RVVM1HI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HI, true, RVVMF2HI, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HI, true, RVVM1HI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HI, true, RVVMF2HI, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HI, true, RVVM1HI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HI, true, RVVMF2HI, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HI, true, RVVM1HI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HI, true, RVVMF2HI, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HI, true, RVVM2HI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HI, true, RVVM1HI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HI, true, RVVMF2HI, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HI, true, RVVM2HI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HI, true, RVVM1HI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HI, true, RVVMF2HI, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HI, true, RVVM4HI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HI, true, RVVM2HI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HI, true, RVVM1HI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HI, true, RVVMF2HI, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM4HF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8SI, true, RVVM1SI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SI, true, RVVM1SI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SI, true, RVVM1SI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SI, true, RVVM1SI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SI, true, RVVM2SI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SI, true, RVVM1SI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SI, true, RVVM2SI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SI, true, RVVM1SI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SI, true, RVVM4SI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SI, true, RVVM2SI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SI, true, RVVM1SI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM4SF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 8, LMUL_1, 16) TUPLE_ENTRY (RVVM1x7DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 7, LMUL_1, 16) diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index 8f74cd2e502..28cf1f414e9 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -1406,6 +1406,9 @@ riscv_v_adjust_bytesize (machine_mode mode, int scale) { if (riscv_v_ext_vector_mode_p (mode)) { + if (TARGET_XTHEADVECTOR) + return BYTES_PER_RISCV_VECTOR; + poly_int64 nunits = GET_MODE_NUNITS (mode); poly_int64 mode_size = GET_MODE_SIZE (mode); @@ -9995,7 +9998,7 @@ riscv_use_divmod_expander (void) static machine_mode riscv_preferred_simd_mode (scalar_mode mode) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::preferred_simd_mode (mode); return word_mode; @@ -10346,7 +10349,7 @@ riscv_mode_priority (int, int n) unsigned int riscv_autovectorize_vector_modes (vector_modes *modes, bool all) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::autovectorize_vector_modes (modes, all); return default_autovectorize_vector_modes (modes, all); @@ -10529,6 +10532,16 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) return false; } +/* Implements target hook vector_mode_supported_any_target_p. */ + +static bool +riscv_vector_mode_supported_any_target_p (machine_mode mode) +{ + if (TARGET_XTHEADVECTOR) + return false; + return true; +} + /* Initialize the GCC target structure. */ #undef TARGET_ASM_ALIGNED_HI_OP #define TARGET_ASM_ALIGNED_HI_OP "\t.half\t" @@ -10872,6 +10885,9 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) #undef TARGET_PREFERRED_ELSE_VALUE #define TARGET_PREFERRED_ELSE_VALUE riscv_preferred_else_value +#undef TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P +#define TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P riscv_vector_mode_supported_any_target_p + struct gcc_target targetm = TARGET_INITIALIZER; #include "gt-riscv.h" diff --git a/gcc/config/riscv/riscv_th_vector.h b/gcc/config/riscv/riscv_th_vector.h new file mode 100644 index 00000000000..b6b6738bdda --- /dev/null +++ b/gcc/config/riscv/riscv_th_vector.h @@ -0,0 +1,49 @@ +/* RISC-V 'XTheadVector' Extension intrinsics include file. + Copyright (C) 2024 Free Software Foundation, Inc. + + This file is part of GCC. + + GCC is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published + by the Free Software Foundation; either version 3, or (at your + option) any later version. + + GCC is distributed in the hope that it will be useful, but WITHOUT + ANY WARRANTY; without even the implied warranty of MERCHANTABILITY + or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public + License for more details. + + Under Section 7 of GPL version 3, you are granted additional + permissions described in the GCC Runtime Library Exception, version + 3.1, as published by the Free Software Foundation. + + You should have received a copy of the GNU General Public License and + a copy of the GCC Runtime Library Exception along with this program; + see the files COPYING3 and COPYING.RUNTIME respectively. If not, see + <" target="_blank">http://www.gnu.org/licenses/>. */ + +#ifndef __RISCV_TH_VECTOR_H +#define __RISCV_TH_VECTOR_H + +#include <stdint.h> +#include <stddef.h> + +#ifndef __riscv_xtheadvector +#error "XTheadVector intrinsics require the xtheadvector extension." +#else +#ifdef __cplusplus +extern "C" { +#endif + +/* NOTE: This implementation of riscv_th_vector.h is intentionally short. It does + not define the RVV types and intrinsic functions directly in C and C++ + code, but instead uses the following pragma to tell GCC to insert the + necessary type and function definitions itself. The net effect is the + same, and the file is a complete implementation of riscv_th_vector.h. */ +#pragma riscv intrinsic "xtheadvector" + +#ifdef __cplusplus +} +#endif // __cplusplus +#endif // __riscv_xtheadvector +#endif // __RISCV_TH_ECTOR_H diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md new file mode 100644 index 00000000000..696b815252d --- /dev/null +++ b/gcc/config/riscv/thead-vector.md @@ -0,0 +1,102 @@ +(define_c_enum "unspec" [ + UNSPEC_TH_VWLDST +]) + +(define_mode_iterator V_VLS_VT [V VLS VT]) +(define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) + +(define_split + [(set (match_operand:V_VB_VLS_VT 0 "reg_or_mem_operand") + (match_operand:V_VB_VLS_VT 1 "reg_or_mem_operand"))] + "TARGET_XTHEADVECTOR" + [(const_int 0)] + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + }) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:V_VLS_VT 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:V_VLS_VT + [(match_operand:V_VLS_VT 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2")]) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:VB 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:VB + [(match_operand:VB 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2") + (set (attr "sew") (const_int 8)) + (set (attr "vlmul") (symbol_ref "riscv_vector::LMUL_1"))]) + +(define_insn_and_split "*pred_th_mov<mode>" + [(set (match_operand:VB_VLS 0 "nonimmediate_operand" "=vr, m, vr, vr, vr") + (if_then_else:VB_VLS + (unspec:VB_VLS + [(match_operand:VB_VLS 1 "vector_all_trues_mask_operand" "Wc1, Wc1, Wc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) + (match_operand:VB_VLS 3 "vector_move_operand" " m, vr, vr, Wc0, Wc1") + (match_operand:VB_VLS 2 "vector_undef_operand" " vu, vu, vu, vu, vu")))] + "TARGET_XTHEADVECTOR" + "@ + # + # + vmcpy.m\t%0,%3 + vmclr.m\t%0 + vmset.m\t%0" + "&& !reload_completed" + [(const_int 0)] + { + if ((MEM_P (operands[0]) || MEM_P (operands[3])) + || (REG_P (operands[0]) && REG_P (operands[3]) + && INTVAL (operands[5]) == riscv_vector::VLMAX)) + { + emit_move_insn (operands[0], operands[3]); + DONE; + } + FAIL; + } + [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/config/riscv/thead.cc b/gcc/config/riscv/thead.cc index 00b2e14c6c0..cc04101b256 100644 --- a/gcc/config/riscv/thead.cc +++ b/gcc/config/riscv/thead.cc @@ -27,6 +27,7 @@ #include "backend.h" #include "tree.h" #include "rtl.h" +#include "insn-attr.h" #include "explow.h" #include "memmodel.h" #include "emit-rtl.h" @@ -890,8 +891,26 @@ th_asm_output_opcode (FILE *asm_out_file, const char *p) { /* We need to add th. prefix to all the xtheadvector instructions here.*/ - if (current_output_insn != NULL && p[0] == 'v') - fputs ("th.", asm_out_file); + if (current_output_insn != NULL) + { + if (get_attr_type (current_output_insn) == TYPE_VSETVL) + { + if (strstr (p, "zero")) + { + if (strstr (p, "zero,zero")) + return "th.vsetvli\tzero,zero,e%0,%m1"; + else + return "th.vsetvli\tzero,%0,e%1,%m2"; + } + else + { + return "th.vsetvli\t%0,%1,e%2,%m3"; + } + } + + if (p[0] == 'v') + fputs ("th.", asm_out_file); + } return p; } diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index 24b7b4394be..3eb6daafbc2 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -87,7 +87,7 @@ ;; check. However, we need default value of SEW for vsetvl instruction since there ;; is no field for ratio in the vsetvl instruction encoding. (define_attr "sew" "" - (cond [(eq_attr "mode" "RVVMF64BI,RVVMF32BI,RVVMF16BI,RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ + (cond [(eq_attr "mode" "RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ RVVM8QI,RVVM4QI,RVVM2QI,RVVM1QI,RVVMF2QI,RVVMF4QI,RVVMF8QI,\ RVVM1x8QI,RVVMF2x8QI,RVVMF4x8QI,RVVMF8x8QI,\ RVVM1x7QI,RVVMF2x7QI,RVVMF4x7QI,RVVMF8x7QI,\ @@ -99,6 +99,18 @@ V1QI,V2QI,V4QI,V8QI,V16QI,V32QI,V64QI,V128QI,V256QI,V512QI,V1024QI,V2048QI,V4096QI,\ V1BI,V2BI,V4BI,V8BI,V16BI,V32BI,V64BI,V128BI,V256BI,V512BI,V1024BI,V2048BI,V4096BI") (const_int 8) + (eq_attr "mode" "RVVMF16BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 16) + (const_int 8)) + (eq_attr "mode" "RVVMF32BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 32) + (const_int 8)) + (eq_attr "mode" "RVVMF64BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 64) + (const_int 8)) (eq_attr "mode" "RVVM8HI,RVVM4HI,RVVM2HI,RVVM1HI,RVVMF2HI,RVVMF4HI,\ RVVM1x8HI,RVVMF2x8HI,RVVMF4x8HI,\ RVVM1x7HI,RVVMF2x7HI,RVVMF4x7HI,\ @@ -159,9 +171,9 @@ (eq_attr "mode" "RVVM4QI,RVVMF2BI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2QI,RVVMF4BI") (symbol_ref "riscv_vector::LMUL_2") (eq_attr "mode" "RVVM1QI,RVVMF8BI") (symbol_ref "riscv_vector::LMUL_1") - (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "riscv_vector::LMUL_F2") - (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "riscv_vector::LMUL_F4") - (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "riscv_vector::LMUL_F8") + (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F2") + (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F4") + (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F8") (eq_attr "mode" "RVVM8HI") (symbol_ref "riscv_vector::LMUL_8") (eq_attr "mode" "RVVM4HI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2HI") (symbol_ref "riscv_vector::LMUL_2") @@ -436,6 +448,10 @@ vaeskf1,vaeskf2,vaesz,vsha2ms,vsha2ch,vsha2cl,vsm4k,vsm4r,\ vsm3me,vsm3c") (const_int INVALID_ATTRIBUTE) + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE) (eq_attr "mode" "RVVM8QI,RVVM1BI") (const_int 1) (eq_attr "mode" "RVVM4QI,RVVMF2BI") (const_int 2) (eq_attr "mode" "RVVM2QI,RVVMF4BI") (const_int 4) @@ -905,6 +921,8 @@ (symbol_ref "riscv_vector::FRM_DYN")] (symbol_ref "riscv_vector::FRM_NONE"))) +(include "thead-vector.md") + ;; ----------------------------------------------------------------- ;; ---- Miscellaneous Operations ;; ----------------------------------------------------------------- @@ -1078,6 +1096,12 @@ - We can not leave it to TARGET_SECONDARY_RELOAD since it happens before spilling. The clobber scratch is used by spilling fractional registers in IRA/LRA so it's too early. */ + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; @@ -1114,7 +1138,7 @@ (define_insn "*mov<mode>_whole" [(set (match_operand:V_WHOLE 0 "reg_or_mem_operand" "=vr, m,vr") (match_operand:V_WHOLE 1 "reg_or_mem_operand" " m,vr,vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "@ vl%m1re<sew>.v\t%0,%1 vs%m1r.v\t%1,%0 @@ -1135,6 +1159,13 @@ (match_operand:VB 1 "general_operand"))] "TARGET_VECTOR" { + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } + if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; }) @@ -1142,7 +1173,7 @@ (define_insn "*mov<mode>" [(set (match_operand:VB 0 "register_operand" "=vr") (match_operand:VB 1 "register_operand" " vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "vmv1r.v\t%0,%1" [(set_attr "type" "vmov") (set_attr "mode" "<MODE>")]) @@ -3692,7 +3723,7 @@ (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3713,7 +3744,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3734,7 +3765,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c index 2e0e12aa045..2eef9e1e1a8 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c @@ -1,4 +1,4 @@ -/* { dg-do compile } */ +/* { dg-do compile { target { ! riscv_xtheadvector } } } */ /* { dg-skip-if "test rvv intrinsic" { *-*-* } { "*" } { "-march=rv*v*" } } */ void foo0 () {__rvv_bool64_t t;} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c index 3d81b179235..ef329e30785 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c @@ -1,4 +1,4 @@ /* { dg-do compile } */ /* { dg-options "-O3 -march=rv32gc -mabi=ilp32d" } */ -#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' extension enabled} } */ +#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' or 'XTHEADVECTOR' extension enabled} } */ diff --git a/gcc/testsuite/lib/target-supports.exp b/gcc/testsuite/lib/target-supports.exp index b27c30b8c51..bf0ad5342d6 100644 --- a/gcc/testsuite/lib/target-supports.exp +++ b/gcc/testsuite/lib/target-supports.exp @@ -1956,6 +1956,18 @@ proc check_effective_target_riscv_zbb { } { }] } +# Return 1 if the target arch supports the XTheadVector extension, 0 otherwise. +# Cache the result. + +proc check_effective_target_riscv_xtheadvector { } { + return [check_no_compiler_messages riscv_ext_xtheadvector assembly { + #ifndef __riscv_xtheadvector + #error "Not __riscv_xtheadvector" + #endif + }] +} + + # Return 1 if we can execute code when using dg-add-options riscv_v proc check_effective_target_riscv_v_ok { } { -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: Re：[PATCH v5] RISC-V: Handle differences between XTheadvector and Vector 2024-01-10 2:57 ` Re：[PATCH " joshua 2024-01-10 3:02 ` juzhe.zhong @ 2024-01-10 6:54 ` juzhe.zhong 2024-01-10 7:01 ` juzhe.zhong 2 siblings, 0 replies; 133+ messages in thread From: juzhe.zhong @ 2024-01-10 6:54 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 48035 bytes --] https://gcc.gnu.org/pipermail/gcc-patches/2024-January/641733.html This patch is ok from my side. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-10 10:57 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Handle differences between XTheadvector and Vector Hi Juzhe, Thank you for so many useful comments for this patch! There are some more patches to support xtheadvector special instrinsics as well as handle register overlap issue and rewrite assembly output. https://gcc.gnu.org/pipermail/gcc-patches/2024-January/641774.html https://gcc.gnu.org/pipermail/gcc-patches/2024-January/641732.html https://gcc.gnu.org/pipermail/gcc-patches/2024-January/641733.html Also, there is a precedent patch to refactor riscv-vector-builtins-bases.cc https://gcc.gnu.org/pipermail/gcc-patches/2023-December/641530.html Jeff has reviewed it, but didn't have LGTM yet. Joshua ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月10日(星期三) 10:34 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Handle differences between XTheadvector and Vector Thanks for your patience. LGTM from myside. I think it's pretty clean now. I can image in the future when some day the theadvector is no longer used, we can remove it very easily. And also, the theadvector won't affect our RVV1.0 maintain since it's isolated cleanly. But I'd like to wait for a few more days some body want to chime in. And you should do more things before commit it: 1. Remember you should run the full coverage RVV1.0 API test, the test-generator is downloaded from official intrinsic doc: https://github.com/riscv-non-isa/rvv-intrinsic-doc 2. Also the regression of RV32 an RV64 of GCC testsuite. Do you have more patches of theadvector that I didn't review ? plz point them to me again. Thanks. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-10 10:22 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Handle differences between XTheadvector and Vector This patch is to handle the differences in instruction generation between Vector and XTheadVector. In this version, we only support partial xtheadvector instructions that leverage directly from current RVV1.0 with simple adding "th." prefix. For different name xtheadvector instructions but share same patterns as RVV1.0 instructions, we will use ASM targethook to rewrite the whole string of the instructions in the following patches. For some vector patterns that cannot be avoided, we use "!TARGET_XTHEADVECTOR" to disable them in vector.md in order not to generate instructions that xtheadvector does not support, like vmv1r and vsext.vf2. gcc/ChangeLog: * config.gcc: Add files for XTheadVector intrinsics. * config/riscv/autovec.md: Guard XTheadVector. * config/riscv/riscv-c.cc: Add pragma for XTheadVector. * config/riscv/riscv-string.cc (expand_block_move): Guard XTheadVector. * config/riscv/riscv-string.cc (vls_mode_valid_p): Avoid autovec. * config/riscv/riscv-vector-builtins-shapes.cc (check_type): (build_one): New function. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): (DEF_THEAD_RVV_FUNCTION): Add new marcos. (check_required_extensions): (handle_pragma_vector): * config/riscv/riscv-vector-builtins.h (RVV_REQUIRE_VECTOR): (RVV_REQUIRE_XTHEADVECTOR): Add RVV_REQUIRE_VECTOR and RVV_REQUIRE_XTHEADVECTOR. (struct function_group_info): * config/riscv/riscv-vector-switch.def (ENTRY): Disable fractional mode for the XTheadVector extension. (TUPLE_ENTRY): Likewise. * config/riscv/riscv-vsetvl.cc: Add functions for xtheadvector. * config/riscv/riscv.cc (riscv_v_ext_vls_mode_p): Guard XTheadVector. (riscv_v_adjust_bytesize): Likewise. (riscv_preferred_simd_mode): Likewsie. (riscv_autovectorize_vector_modes): Likewise. (riscv_vector_mode_supported_any_target_p): Likewise. (TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P): Likewise. * config/riscv/vector.md: Include thead-vector.md. * config/riscv/riscv_th_vector.h: New file. * config/riscv/thead-vector.md: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/base/pragma-1.c: Add XTheadVector. * gcc.target/riscv/rvv/base/abi-1.c: Exclude XTheadVector. * lib/target-supports.exp: Add target for XTheadVector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config.gcc | 2 +- gcc/config/riscv/autovec.md | 2 +- gcc/config/riscv/predicates.md | 4 +- gcc/config/riscv/riscv-c.cc | 3 +- gcc/config/riscv/riscv-string.cc | 3 +- gcc/config/riscv/riscv-v.cc | 2 +- .../riscv/riscv-vector-builtins-bases.cc | 48 ++++-- .../riscv/riscv-vector-builtins-shapes.cc | 23 +++ gcc/config/riscv/riscv-vector-switch.def | 150 +++++++++--------- gcc/config/riscv/riscv.cc | 20 ++- gcc/config/riscv/riscv_th_vector.h | 49 ++++++ gcc/config/riscv/thead-vector.md | 102 ++++++++++++ gcc/config/riscv/thead.cc | 23 ++- gcc/config/riscv/vector.md | 49 ++++-- .../gcc.target/riscv/rvv/base/abi-1.c | 2 +- .../gcc.target/riscv/rvv/base/pragma-1.c | 2 +- gcc/testsuite/lib/target-supports.exp | 12 ++ 17 files changed, 383 insertions(+), 113 deletions(-) create mode 100644 gcc/config/riscv/riscv_th_vector.h create mode 100644 gcc/config/riscv/thead-vector.md diff --git a/gcc/config.gcc b/gcc/config.gcc index 7e583390024..047e4c02cf4 100644 --- a/gcc/config.gcc +++ b/gcc/config.gcc @@ -549,7 +549,7 @@ riscv*) extra_objs="${extra_objs} riscv-vector-builtins.o riscv-vector-builtins-shapes.o riscv-vector-builtins-bases.o" extra_objs="${extra_objs} thead.o riscv-target-attr.o" d_target_objs="riscv-d.o" - extra_headers="riscv_vector.h" + extra_headers="riscv_vector.h riscv_th_vector.h" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.cc" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.h" ;; diff --git a/gcc/config/riscv/autovec.md b/gcc/config/riscv/autovec.md index 775eaa825b0..0477781cabe 100644 --- a/gcc/config/riscv/autovec.md +++ b/gcc/config/riscv/autovec.md @@ -2579,7 +2579,7 @@ [(match_operand 0 "register_operand") (match_operand 1 "memory_operand") (match_operand:ANYI 2 "const_int_operand")] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" { riscv_vector::expand_rawmemchr(<MODE>mode, operands[0], operands[1], operands[2]); diff --git a/gcc/config/riscv/predicates.md b/gcc/config/riscv/predicates.md index b1a79cae50a..0337da88284 100644 --- a/gcc/config/riscv/predicates.md +++ b/gcc/config/riscv/predicates.md @@ -428,7 +428,9 @@ ;; Predicates for the V extension. (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (ior (match_test "TARGET_XTHEADVECTOR && rtx_equal_p (op, const0_rtx)") + (match_test "!TARGET_XTHEADVECTOR")) + (match_operand 0 "const_csr_operand")))) (define_special_predicate "autovec_length_operand" (ior (match_operand 0 "pmode_register_operand") diff --git a/gcc/config/riscv/riscv-c.cc b/gcc/config/riscv/riscv-c.cc index 422ddc2c308..2e306057347 100644 --- a/gcc/config/riscv/riscv-c.cc +++ b/gcc/config/riscv/riscv-c.cc @@ -195,7 +195,8 @@ riscv_pragma_intrinsic (cpp_reader *) const char *name = TREE_STRING_POINTER (x); - if (strcmp (name, "vector") == 0) + if (strcmp (name, "vector") == 0 + || strcmp (name, "xtheadvector") == 0) { if (!TARGET_VECTOR) { diff --git a/gcc/config/riscv/riscv-string.cc b/gcc/config/riscv/riscv-string.cc index f227b8667ce..b09b51d7526 100644 --- a/gcc/config/riscv/riscv-string.cc +++ b/gcc/config/riscv/riscv-string.cc @@ -773,7 +773,8 @@ riscv_expand_block_move_scalar (rtx dest, rtx src, rtx length) bool riscv_expand_block_move (rtx dest, rtx src, rtx length) { - if (TARGET_VECTOR && stringop_strategy & STRATEGY_VECTOR) + if ((TARGET_VECTOR && !TARGET_XTHEADVECTOR) + && stringop_strategy & STRATEGY_VECTOR) { bool ok = riscv_vector::expand_block_move (dest, src, length); if (ok) diff --git a/gcc/config/riscv/riscv-v.cc b/gcc/config/riscv/riscv-v.cc index 2491522191a..7e25deb966d 100644 --- a/gcc/config/riscv/riscv-v.cc +++ b/gcc/config/riscv/riscv-v.cc @@ -4392,7 +4392,7 @@ cmp_lmul_gt_one (machine_mode mode) bool vls_mode_valid_p (machine_mode vls_mode) { - if (!TARGET_VECTOR) + if (!TARGET_VECTOR || TARGET_XTHEADVECTOR) return false; if (riscv_autovec_preference == RVV_SCALABLE) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 91af2c00674..46f1a1da33e 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -115,23 +115,37 @@ public: tree type = builtin_types[e.type.index].vector; machine_mode mode = TYPE_MODE (type); - /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. - - - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 - - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 - - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 - - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 - - e8,m2/e16,m4/e32,m8 --> e8m2 - - e8,m4/e16,m8 --> e8m4 - - e8,m8 --> e8m8 - */ - /* SEW. */ - e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); - - /* LMUL. */ - machine_mode e8_mode - = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); - e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + + if (TARGET_XTHEADVECTOR) + { + machine_mode inner_mode = GET_MODE_INNER (mode); + /* SEW. */ + e.add_input_operand (Pmode, + gen_int_mode (GET_MODE_BITSIZE (inner_mode), Pmode)); + /* LMUL. */ + e.add_input_operand (Pmode, + gen_int_mode (get_vlmul (mode), Pmode)); + } + else + { + /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. + + - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 + - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 + - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 + - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 + - e8,m2/e16,m4/e32,m8 --> e8m2 + - e8,m4/e16,m8 --> e8m4 + - e8,m8 --> e8m8 + */ + /* SEW. */ + e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); + + /* LMUL. */ + machine_mode e8_mode + = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); + e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + } /* TAIL_ANY. */ e.add_input_operand (Pmode, diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index ee8058dcdb1..1e4f4d53de6 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -33,6 +33,25 @@ namespace riscv_vector { +/* Check whether the RETURN_TYPE and ARGUMENT_TYPES are + valid for the function. */ + +static bool +check_type (tree return_type, vec<tree> &argument_types) +{ + tree arg; + unsigned i; + + if (!return_type) + return false; + + FOR_EACH_VEC_ELT (argument_types, i, arg) + if (!arg) + return false; + + return true; +} + /* Add one function instance for GROUP, using operand suffix at index OI, mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ static void @@ -49,6 +68,10 @@ build_one (function_builder &b, const function_group_info &group, group.ops_infos.types[vec_type_idx].index); b.allocate_argument_types (function_instance, argument_types); b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + b.add_overloaded_function (function_instance, *group.shape); b.add_unique_function (function_instance, (*group.shape), return_type, argument_types); diff --git a/gcc/config/riscv/riscv-vector-switch.def b/gcc/config/riscv/riscv-vector-switch.def index 1ad26c2a3b2..452283b7416 100644 --- a/gcc/config/riscv/riscv-vector-switch.def +++ b/gcc/config/riscv/riscv-vector-switch.def @@ -68,9 +68,9 @@ Encode the ratio of SEW/LMUL into the mask types. #endif /* Disable modes if TARGET_MIN_VLEN == 32. */ -ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) -ENTRY (RVVMF32BI, true, LMUL_F4, 32) -ENTRY (RVVMF16BI, true, LMUL_F2, 16) +ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F8, 64) +ENTRY (RVVMF32BI, true, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F4, 32) +ENTRY (RVVMF16BI, true, TARGET_XTHEADVECTOR ? LMUL_1 : LMUL_F2 , 16) ENTRY (RVVMF8BI, true, LMUL_1, 8) ENTRY (RVVMF4BI, true, LMUL_2, 4) ENTRY (RVVMF2BI, true, LMUL_4, 2) @@ -81,39 +81,39 @@ ENTRY (RVVM8QI, true, LMUL_8, 1) ENTRY (RVVM4QI, true, LMUL_4, 2) ENTRY (RVVM2QI, true, LMUL_2, 4) ENTRY (RVVM1QI, true, LMUL_1, 8) -ENTRY (RVVMF2QI, true, LMUL_F2, 16) -ENTRY (RVVMF4QI, true, LMUL_F4, 32) -ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) +ENTRY (RVVMF2QI, !TARGET_XTHEADVECTOR, LMUL_F2, 16) +ENTRY (RVVMF4QI, !TARGET_XTHEADVECTOR, LMUL_F4, 32) +ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F8, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8HI, true, LMUL_8, 2) ENTRY (RVVM4HI, true, LMUL_4, 4) ENTRY (RVVM2HI, true, LMUL_2, 8) ENTRY (RVVM1HI, true, LMUL_1, 16) -ENTRY (RVVMF2HI, true, LMUL_F2, 32) -ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HI, !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_16. */ ENTRY (RVVM8HF, TARGET_VECTOR_ELEN_FP_16, LMUL_8, 2) ENTRY (RVVM4HF, TARGET_VECTOR_ELEN_FP_16, LMUL_4, 4) ENTRY (RVVM2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_2, 8) ENTRY (RVVM1HF, TARGET_VECTOR_ELEN_FP_16, LMUL_1, 16) -ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_F2, 32) -ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8SI, true, LMUL_8, 4) ENTRY (RVVM4SI, true, LMUL_4, 8) ENTRY (RVVM2SI, true, LMUL_2, 16) ENTRY (RVVM1SI, true, LMUL_1, 32) -ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_32. */ ENTRY (RVVM8SF, TARGET_VECTOR_ELEN_FP_32, LMUL_8, 4) ENTRY (RVVM4SF, TARGET_VECTOR_ELEN_FP_32, LMUL_4, 8) ENTRY (RVVM2SF, TARGET_VECTOR_ELEN_FP_32, LMUL_2, 16) ENTRY (RVVM1SF, TARGET_VECTOR_ELEN_FP_32, LMUL_1, 32) -ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if !TARGET_VECTOR_ELEN_64. */ ENTRY (RVVM8DI, TARGET_VECTOR_ELEN_64, LMUL_8, 8) @@ -140,127 +140,127 @@ ENTRY (RVVM1DF, TARGET_VECTOR_ELEN_FP_64, LMUL_1, 64) #endif TUPLE_ENTRY (RVVM1x8QI, true, RVVM1QI, 8, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x8QI, true, RVVMF2QI, 8, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x8QI, true, RVVMF4QI, 8, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 8, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x8QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 8, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x8QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 8, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 8, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x7QI, true, RVVM1QI, 7, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x7QI, true, RVVMF2QI, 7, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x7QI, true, RVVMF4QI, 7, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 7, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x7QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 7, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x7QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 7, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 7, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x6QI, true, RVVM1QI, 6, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x6QI, true, RVVMF2QI, 6, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x6QI, true, RVVMF4QI, 6, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 6, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x6QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 6, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x6QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 6, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 6, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x5QI, true, RVVM1QI, 5, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x5QI, true, RVVMF2QI, 5, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x5QI, true, RVVMF4QI, 5, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 5, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x5QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 5, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x5QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 5, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 5, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x4QI, true, RVVM2QI, 4, LMUL_2, 4) TUPLE_ENTRY (RVVM1x4QI, true, RVVM1QI, 4, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x4QI, true, RVVMF2QI, 4, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x4QI, true, RVVMF4QI, 4, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 4, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x4QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 4, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x4QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 4, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 4, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x3QI, true, RVVM2QI, 3, LMUL_2, 4) TUPLE_ENTRY (RVVM1x3QI, true, RVVM1QI, 3, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x3QI, true, RVVMF2QI, 3, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x3QI, true, RVVMF4QI, 3, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 3, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x3QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 3, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x3QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 3, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 3, LMUL_F8, 64) TUPLE_ENTRY (RVVM4x2QI, true, RVVM4QI, 2, LMUL_4, 2) TUPLE_ENTRY (RVVM2x2QI, true, RVVM2QI, 2, LMUL_2, 4) TUPLE_ENTRY (RVVM1x2QI, true, RVVM1QI, 2, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x2QI, true, RVVMF2QI, 2, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x2QI, true, RVVMF4QI, 2, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 2, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x2QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 2, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x2QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 2, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 2, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x8HI, true, RVVM1HI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HI, true, RVVMF2HI, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HI, true, RVVM1HI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HI, true, RVVMF2HI, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HI, true, RVVM1HI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HI, true, RVVMF2HI, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HI, true, RVVM1HI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HI, true, RVVMF2HI, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HI, true, RVVM2HI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HI, true, RVVM1HI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HI, true, RVVMF2HI, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HI, true, RVVM2HI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HI, true, RVVM1HI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HI, true, RVVMF2HI, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HI, true, RVVM4HI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HI, true, RVVM2HI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HI, true, RVVM1HI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HI, true, RVVMF2HI, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM4HF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8SI, true, RVVM1SI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SI, true, RVVM1SI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SI, true, RVVM1SI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SI, true, RVVM1SI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SI, true, RVVM2SI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SI, true, RVVM1SI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SI, true, RVVM2SI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SI, true, RVVM1SI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SI, true, RVVM4SI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SI, true, RVVM2SI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SI, true, RVVM1SI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM4SF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 8, LMUL_1, 16) TUPLE_ENTRY (RVVM1x7DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 7, LMUL_1, 16) diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index 8f74cd2e502..28cf1f414e9 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -1406,6 +1406,9 @@ riscv_v_adjust_bytesize (machine_mode mode, int scale) { if (riscv_v_ext_vector_mode_p (mode)) { + if (TARGET_XTHEADVECTOR) + return BYTES_PER_RISCV_VECTOR; + poly_int64 nunits = GET_MODE_NUNITS (mode); poly_int64 mode_size = GET_MODE_SIZE (mode); @@ -9995,7 +9998,7 @@ riscv_use_divmod_expander (void) static machine_mode riscv_preferred_simd_mode (scalar_mode mode) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::preferred_simd_mode (mode); return word_mode; @@ -10346,7 +10349,7 @@ riscv_mode_priority (int, int n) unsigned int riscv_autovectorize_vector_modes (vector_modes *modes, bool all) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::autovectorize_vector_modes (modes, all); return default_autovectorize_vector_modes (modes, all); @@ -10529,6 +10532,16 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) return false; } +/* Implements target hook vector_mode_supported_any_target_p. */ + +static bool +riscv_vector_mode_supported_any_target_p (machine_mode mode) +{ + if (TARGET_XTHEADVECTOR) + return false; + return true; +} + /* Initialize the GCC target structure. */ #undef TARGET_ASM_ALIGNED_HI_OP #define TARGET_ASM_ALIGNED_HI_OP "\t.half\t" @@ -10872,6 +10885,9 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) #undef TARGET_PREFERRED_ELSE_VALUE #define TARGET_PREFERRED_ELSE_VALUE riscv_preferred_else_value +#undef TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P +#define TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P riscv_vector_mode_supported_any_target_p + struct gcc_target targetm = TARGET_INITIALIZER; #include "gt-riscv.h" diff --git a/gcc/config/riscv/riscv_th_vector.h b/gcc/config/riscv/riscv_th_vector.h new file mode 100644 index 00000000000..b6b6738bdda --- /dev/null +++ b/gcc/config/riscv/riscv_th_vector.h @@ -0,0 +1,49 @@ +/* RISC-V 'XTheadVector' Extension intrinsics include file. + Copyright (C) 2024 Free Software Foundation, Inc. + + This file is part of GCC. + + GCC is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published + by the Free Software Foundation; either version 3, or (at your + option) any later version. + + GCC is distributed in the hope that it will be useful, but WITHOUT + ANY WARRANTY; without even the implied warranty of MERCHANTABILITY + or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public + License for more details. + + Under Section 7 of GPL version 3, you are granted additional + permissions described in the GCC Runtime Library Exception, version + 3.1, as published by the Free Software Foundation. + + You should have received a copy of the GNU General Public License and + a copy of the GCC Runtime Library Exception along with this program; + see the files COPYING3 and COPYING.RUNTIME respectively. If not, see + <" target="_blank">http://www.gnu.org/licenses/>. */ + +#ifndef __RISCV_TH_VECTOR_H +#define __RISCV_TH_VECTOR_H + +#include <stdint.h> +#include <stddef.h> + +#ifndef __riscv_xtheadvector +#error "XTheadVector intrinsics require the xtheadvector extension." +#else +#ifdef __cplusplus +extern "C" { +#endif + +/* NOTE: This implementation of riscv_th_vector.h is intentionally short. It does + not define the RVV types and intrinsic functions directly in C and C++ + code, but instead uses the following pragma to tell GCC to insert the + necessary type and function definitions itself. The net effect is the + same, and the file is a complete implementation of riscv_th_vector.h. */ +#pragma riscv intrinsic "xtheadvector" + +#ifdef __cplusplus +} +#endif // __cplusplus +#endif // __riscv_xtheadvector +#endif // __RISCV_TH_ECTOR_H diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md new file mode 100644 index 00000000000..696b815252d --- /dev/null +++ b/gcc/config/riscv/thead-vector.md @@ -0,0 +1,102 @@ +(define_c_enum "unspec" [ + UNSPEC_TH_VWLDST +]) + +(define_mode_iterator V_VLS_VT [V VLS VT]) +(define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) + +(define_split + [(set (match_operand:V_VB_VLS_VT 0 "reg_or_mem_operand") + (match_operand:V_VB_VLS_VT 1 "reg_or_mem_operand"))] + "TARGET_XTHEADVECTOR" + [(const_int 0)] + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + }) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:V_VLS_VT 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:V_VLS_VT + [(match_operand:V_VLS_VT 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2")]) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:VB 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:VB + [(match_operand:VB 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2") + (set (attr "sew") (const_int 8)) + (set (attr "vlmul") (symbol_ref "riscv_vector::LMUL_1"))]) + +(define_insn_and_split "*pred_th_mov<mode>" + [(set (match_operand:VB_VLS 0 "nonimmediate_operand" "=vr, m, vr, vr, vr") + (if_then_else:VB_VLS + (unspec:VB_VLS + [(match_operand:VB_VLS 1 "vector_all_trues_mask_operand" "Wc1, Wc1, Wc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) + (match_operand:VB_VLS 3 "vector_move_operand" " m, vr, vr, Wc0, Wc1") + (match_operand:VB_VLS 2 "vector_undef_operand" " vu, vu, vu, vu, vu")))] + "TARGET_XTHEADVECTOR" + "@ + # + # + vmcpy.m\t%0,%3 + vmclr.m\t%0 + vmset.m\t%0" + "&& !reload_completed" + [(const_int 0)] + { + if ((MEM_P (operands[0]) || MEM_P (operands[3])) + || (REG_P (operands[0]) && REG_P (operands[3]) + && INTVAL (operands[5]) == riscv_vector::VLMAX)) + { + emit_move_insn (operands[0], operands[3]); + DONE; + } + FAIL; + } + [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/config/riscv/thead.cc b/gcc/config/riscv/thead.cc index 00b2e14c6c0..cc04101b256 100644 --- a/gcc/config/riscv/thead.cc +++ b/gcc/config/riscv/thead.cc @@ -27,6 +27,7 @@ #include "backend.h" #include "tree.h" #include "rtl.h" +#include "insn-attr.h" #include "explow.h" #include "memmodel.h" #include "emit-rtl.h" @@ -890,8 +891,26 @@ th_asm_output_opcode (FILE *asm_out_file, const char *p) { /* We need to add th. prefix to all the xtheadvector instructions here.*/ - if (current_output_insn != NULL && p[0] == 'v') - fputs ("th.", asm_out_file); + if (current_output_insn != NULL) + { + if (get_attr_type (current_output_insn) == TYPE_VSETVL) + { + if (strstr (p, "zero")) + { + if (strstr (p, "zero,zero")) + return "th.vsetvli\tzero,zero,e%0,%m1"; + else + return "th.vsetvli\tzero,%0,e%1,%m2"; + } + else + { + return "th.vsetvli\t%0,%1,e%2,%m3"; + } + } + + if (p[0] == 'v') + fputs ("th.", asm_out_file); + } return p; } diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index 24b7b4394be..3eb6daafbc2 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -87,7 +87,7 @@ ;; check. However, we need default value of SEW for vsetvl instruction since there ;; is no field for ratio in the vsetvl instruction encoding. (define_attr "sew" "" - (cond [(eq_attr "mode" "RVVMF64BI,RVVMF32BI,RVVMF16BI,RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ + (cond [(eq_attr "mode" "RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ RVVM8QI,RVVM4QI,RVVM2QI,RVVM1QI,RVVMF2QI,RVVMF4QI,RVVMF8QI,\ RVVM1x8QI,RVVMF2x8QI,RVVMF4x8QI,RVVMF8x8QI,\ RVVM1x7QI,RVVMF2x7QI,RVVMF4x7QI,RVVMF8x7QI,\ @@ -99,6 +99,18 @@ V1QI,V2QI,V4QI,V8QI,V16QI,V32QI,V64QI,V128QI,V256QI,V512QI,V1024QI,V2048QI,V4096QI,\ V1BI,V2BI,V4BI,V8BI,V16BI,V32BI,V64BI,V128BI,V256BI,V512BI,V1024BI,V2048BI,V4096BI") (const_int 8) + (eq_attr "mode" "RVVMF16BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 16) + (const_int 8)) + (eq_attr "mode" "RVVMF32BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 32) + (const_int 8)) + (eq_attr "mode" "RVVMF64BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 64) + (const_int 8)) (eq_attr "mode" "RVVM8HI,RVVM4HI,RVVM2HI,RVVM1HI,RVVMF2HI,RVVMF4HI,\ RVVM1x8HI,RVVMF2x8HI,RVVMF4x8HI,\ RVVM1x7HI,RVVMF2x7HI,RVVMF4x7HI,\ @@ -159,9 +171,9 @@ (eq_attr "mode" "RVVM4QI,RVVMF2BI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2QI,RVVMF4BI") (symbol_ref "riscv_vector::LMUL_2") (eq_attr "mode" "RVVM1QI,RVVMF8BI") (symbol_ref "riscv_vector::LMUL_1") - (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "riscv_vector::LMUL_F2") - (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "riscv_vector::LMUL_F4") - (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "riscv_vector::LMUL_F8") + (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F2") + (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F4") + (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F8") (eq_attr "mode" "RVVM8HI") (symbol_ref "riscv_vector::LMUL_8") (eq_attr "mode" "RVVM4HI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2HI") (symbol_ref "riscv_vector::LMUL_2") @@ -436,6 +448,10 @@ vaeskf1,vaeskf2,vaesz,vsha2ms,vsha2ch,vsha2cl,vsm4k,vsm4r,\ vsm3me,vsm3c") (const_int INVALID_ATTRIBUTE) + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE) (eq_attr "mode" "RVVM8QI,RVVM1BI") (const_int 1) (eq_attr "mode" "RVVM4QI,RVVMF2BI") (const_int 2) (eq_attr "mode" "RVVM2QI,RVVMF4BI") (const_int 4) @@ -905,6 +921,8 @@ (symbol_ref "riscv_vector::FRM_DYN")] (symbol_ref "riscv_vector::FRM_NONE"))) +(include "thead-vector.md") + ;; ----------------------------------------------------------------- ;; ---- Miscellaneous Operations ;; ----------------------------------------------------------------- @@ -1078,6 +1096,12 @@ - We can not leave it to TARGET_SECONDARY_RELOAD since it happens before spilling. The clobber scratch is used by spilling fractional registers in IRA/LRA so it's too early. */ + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; @@ -1114,7 +1138,7 @@ (define_insn "*mov<mode>_whole" [(set (match_operand:V_WHOLE 0 "reg_or_mem_operand" "=vr, m,vr") (match_operand:V_WHOLE 1 "reg_or_mem_operand" " m,vr,vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "@ vl%m1re<sew>.v\t%0,%1 vs%m1r.v\t%1,%0 @@ -1135,6 +1159,13 @@ (match_operand:VB 1 "general_operand"))] "TARGET_VECTOR" { + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } + if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; }) @@ -1142,7 +1173,7 @@ (define_insn "*mov<mode>" [(set (match_operand:VB 0 "register_operand" "=vr") (match_operand:VB 1 "register_operand" " vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "vmv1r.v\t%0,%1" [(set_attr "type" "vmov") (set_attr "mode" "<MODE>")]) @@ -3692,7 +3723,7 @@ (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3713,7 +3744,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3734,7 +3765,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c index 2e0e12aa045..2eef9e1e1a8 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c @@ -1,4 +1,4 @@ -/* { dg-do compile } */ +/* { dg-do compile { target { ! riscv_xtheadvector } } } */ /* { dg-skip-if "test rvv intrinsic" { *-*-* } { "*" } { "-march=rv*v*" } } */ void foo0 () {__rvv_bool64_t t;} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c index 3d81b179235..ef329e30785 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c @@ -1,4 +1,4 @@ /* { dg-do compile } */ /* { dg-options "-O3 -march=rv32gc -mabi=ilp32d" } */ -#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' extension enabled} } */ +#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' or 'XTHEADVECTOR' extension enabled} } */ diff --git a/gcc/testsuite/lib/target-supports.exp b/gcc/testsuite/lib/target-supports.exp index b27c30b8c51..bf0ad5342d6 100644 --- a/gcc/testsuite/lib/target-supports.exp +++ b/gcc/testsuite/lib/target-supports.exp @@ -1956,6 +1956,18 @@ proc check_effective_target_riscv_zbb { } { }] } +# Return 1 if the target arch supports the XTheadVector extension, 0 otherwise. +# Cache the result. + +proc check_effective_target_riscv_xtheadvector { } { + return [check_no_compiler_messages riscv_ext_xtheadvector assembly { + #ifndef __riscv_xtheadvector + #error "Not __riscv_xtheadvector" + #endif + }] +} + + # Return 1 if we can execute code when using dg-add-options riscv_v proc check_effective_target_riscv_v_ok { } { -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: Re：[PATCH v5] RISC-V: Handle differences between XTheadvector and Vector 2024-01-10 2:57 ` Re：[PATCH " joshua 2024-01-10 3:02 ` juzhe.zhong 2024-01-10 6:54 ` juzhe.zhong @ 2024-01-10 7:01 ` juzhe.zhong 2024-01-10 7:16 ` Re：Re：[PATCH " joshua 2 siblings, 1 reply; 133+ messages in thread From: juzhe.zhong @ 2024-01-10 7:01 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 48851 bytes --] Why do you add theadvector shapes ? I think you can reuse the current existing shapes. +thead-vector-builtins.o: \ + $(srcdir)/config/riscv/thead-vector-builtins.cc \ + $(CONFIG_H) $(SYSTEM_H) coretypes.h $(TM_H) $(TREE_H) $(RTL_H) \ + $(TM_P_H) memmodel.h insn-codes.h $(OPTABS_H) $(RECOG_H) \ + $(EXPR_H) $(BASIC_BLOCK_H) $(FUNCTION_H) fold-const.h $(GIMPLE_H) \ + gimple-iterator.h gimplify.h explow.h $(EMIT_RTL_H) tree-vector-builder.h \ + rtx-vector-builder.h \ + $(srcdir)/config/riscv/riscv-vector-builtins-shapes.h \ + $(srcdir)/config/riscv/riscv-vector-builtins-bases.h \ + $(srcdir)/config/riscv/thead-vector-builtins.h \ + $(RISCV_BUILTINS_H) + $(COMPILER) -c $(ALL_COMPILERFLAGS) $(ALL_CPPFLAGS) $(INCLUDES) \ + $(srcdir)/config/riscv/thead-vector-builtins.cc + Why do you rebuild another new object ? + Copyright (C) 2022-2023 Free Software Foundation, Inc. Incorrect copyright juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-10 10:57 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Handle differences between XTheadvector and Vector Hi Juzhe, Thank you for so many useful comments for this patch! There are some more patches to support xtheadvector special instrinsics as well as handle register overlap issue and rewrite assembly output. https://gcc.gnu.org/pipermail/gcc-patches/2024-January/641774.html https://gcc.gnu.org/pipermail/gcc-patches/2024-January/641732.html https://gcc.gnu.org/pipermail/gcc-patches/2024-January/641733.html Also, there is a precedent patch to refactor riscv-vector-builtins-bases.cc https://gcc.gnu.org/pipermail/gcc-patches/2023-December/641530.html Jeff has reviewed it, but didn't have LGTM yet. Joshua ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月10日(星期三) 10:34 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Handle differences between XTheadvector and Vector Thanks for your patience. LGTM from myside. I think it's pretty clean now. I can image in the future when some day the theadvector is no longer used, we can remove it very easily. And also, the theadvector won't affect our RVV1.0 maintain since it's isolated cleanly. But I'd like to wait for a few more days some body want to chime in. And you should do more things before commit it: 1. Remember you should run the full coverage RVV1.0 API test, the test-generator is downloaded from official intrinsic doc: https://github.com/riscv-non-isa/rvv-intrinsic-doc 2. Also the regression of RV32 an RV64 of GCC testsuite. Do you have more patches of theadvector that I didn't review ? plz point them to me again. Thanks. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-10 10:22 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Handle differences between XTheadvector and Vector This patch is to handle the differences in instruction generation between Vector and XTheadVector. In this version, we only support partial xtheadvector instructions that leverage directly from current RVV1.0 with simple adding "th." prefix. For different name xtheadvector instructions but share same patterns as RVV1.0 instructions, we will use ASM targethook to rewrite the whole string of the instructions in the following patches. For some vector patterns that cannot be avoided, we use "!TARGET_XTHEADVECTOR" to disable them in vector.md in order not to generate instructions that xtheadvector does not support, like vmv1r and vsext.vf2. gcc/ChangeLog: * config.gcc: Add files for XTheadVector intrinsics. * config/riscv/autovec.md: Guard XTheadVector. * config/riscv/riscv-c.cc: Add pragma for XTheadVector. * config/riscv/riscv-string.cc (expand_block_move): Guard XTheadVector. * config/riscv/riscv-string.cc (vls_mode_valid_p): Avoid autovec. * config/riscv/riscv-vector-builtins-shapes.cc (check_type): (build_one): New function. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): (DEF_THEAD_RVV_FUNCTION): Add new marcos. (check_required_extensions): (handle_pragma_vector): * config/riscv/riscv-vector-builtins.h (RVV_REQUIRE_VECTOR): (RVV_REQUIRE_XTHEADVECTOR): Add RVV_REQUIRE_VECTOR and RVV_REQUIRE_XTHEADVECTOR. (struct function_group_info): * config/riscv/riscv-vector-switch.def (ENTRY): Disable fractional mode for the XTheadVector extension. (TUPLE_ENTRY): Likewise. * config/riscv/riscv-vsetvl.cc: Add functions for xtheadvector. * config/riscv/riscv.cc (riscv_v_ext_vls_mode_p): Guard XTheadVector. (riscv_v_adjust_bytesize): Likewise. (riscv_preferred_simd_mode): Likewsie. (riscv_autovectorize_vector_modes): Likewise. (riscv_vector_mode_supported_any_target_p): Likewise. (TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P): Likewise. * config/riscv/vector.md: Include thead-vector.md. * config/riscv/riscv_th_vector.h: New file. * config/riscv/thead-vector.md: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/base/pragma-1.c: Add XTheadVector. * gcc.target/riscv/rvv/base/abi-1.c: Exclude XTheadVector. * lib/target-supports.exp: Add target for XTheadVector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config.gcc | 2 +- gcc/config/riscv/autovec.md | 2 +- gcc/config/riscv/predicates.md | 4 +- gcc/config/riscv/riscv-c.cc | 3 +- gcc/config/riscv/riscv-string.cc | 3 +- gcc/config/riscv/riscv-v.cc | 2 +- .../riscv/riscv-vector-builtins-bases.cc | 48 ++++-- .../riscv/riscv-vector-builtins-shapes.cc | 23 +++ gcc/config/riscv/riscv-vector-switch.def | 150 +++++++++--------- gcc/config/riscv/riscv.cc | 20 ++- gcc/config/riscv/riscv_th_vector.h | 49 ++++++ gcc/config/riscv/thead-vector.md | 102 ++++++++++++ gcc/config/riscv/thead.cc | 23 ++- gcc/config/riscv/vector.md | 49 ++++-- .../gcc.target/riscv/rvv/base/abi-1.c | 2 +- .../gcc.target/riscv/rvv/base/pragma-1.c | 2 +- gcc/testsuite/lib/target-supports.exp | 12 ++ 17 files changed, 383 insertions(+), 113 deletions(-) create mode 100644 gcc/config/riscv/riscv_th_vector.h create mode 100644 gcc/config/riscv/thead-vector.md diff --git a/gcc/config.gcc b/gcc/config.gcc index 7e583390024..047e4c02cf4 100644 --- a/gcc/config.gcc +++ b/gcc/config.gcc @@ -549,7 +549,7 @@ riscv*) extra_objs="${extra_objs} riscv-vector-builtins.o riscv-vector-builtins-shapes.o riscv-vector-builtins-bases.o" extra_objs="${extra_objs} thead.o riscv-target-attr.o" d_target_objs="riscv-d.o" - extra_headers="riscv_vector.h" + extra_headers="riscv_vector.h riscv_th_vector.h" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.cc" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.h" ;; diff --git a/gcc/config/riscv/autovec.md b/gcc/config/riscv/autovec.md index 775eaa825b0..0477781cabe 100644 --- a/gcc/config/riscv/autovec.md +++ b/gcc/config/riscv/autovec.md @@ -2579,7 +2579,7 @@ [(match_operand 0 "register_operand") (match_operand 1 "memory_operand") (match_operand:ANYI 2 "const_int_operand")] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" { riscv_vector::expand_rawmemchr(<MODE>mode, operands[0], operands[1], operands[2]); diff --git a/gcc/config/riscv/predicates.md b/gcc/config/riscv/predicates.md index b1a79cae50a..0337da88284 100644 --- a/gcc/config/riscv/predicates.md +++ b/gcc/config/riscv/predicates.md @@ -428,7 +428,9 @@ ;; Predicates for the V extension. (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (ior (match_test "TARGET_XTHEADVECTOR && rtx_equal_p (op, const0_rtx)") + (match_test "!TARGET_XTHEADVECTOR")) + (match_operand 0 "const_csr_operand")))) (define_special_predicate "autovec_length_operand" (ior (match_operand 0 "pmode_register_operand") diff --git a/gcc/config/riscv/riscv-c.cc b/gcc/config/riscv/riscv-c.cc index 422ddc2c308..2e306057347 100644 --- a/gcc/config/riscv/riscv-c.cc +++ b/gcc/config/riscv/riscv-c.cc @@ -195,7 +195,8 @@ riscv_pragma_intrinsic (cpp_reader *) const char *name = TREE_STRING_POINTER (x); - if (strcmp (name, "vector") == 0) + if (strcmp (name, "vector") == 0 + || strcmp (name, "xtheadvector") == 0) { if (!TARGET_VECTOR) { diff --git a/gcc/config/riscv/riscv-string.cc b/gcc/config/riscv/riscv-string.cc index f227b8667ce..b09b51d7526 100644 --- a/gcc/config/riscv/riscv-string.cc +++ b/gcc/config/riscv/riscv-string.cc @@ -773,7 +773,8 @@ riscv_expand_block_move_scalar (rtx dest, rtx src, rtx length) bool riscv_expand_block_move (rtx dest, rtx src, rtx length) { - if (TARGET_VECTOR && stringop_strategy & STRATEGY_VECTOR) + if ((TARGET_VECTOR && !TARGET_XTHEADVECTOR) + && stringop_strategy & STRATEGY_VECTOR) { bool ok = riscv_vector::expand_block_move (dest, src, length); if (ok) diff --git a/gcc/config/riscv/riscv-v.cc b/gcc/config/riscv/riscv-v.cc index 2491522191a..7e25deb966d 100644 --- a/gcc/config/riscv/riscv-v.cc +++ b/gcc/config/riscv/riscv-v.cc @@ -4392,7 +4392,7 @@ cmp_lmul_gt_one (machine_mode mode) bool vls_mode_valid_p (machine_mode vls_mode) { - if (!TARGET_VECTOR) + if (!TARGET_VECTOR || TARGET_XTHEADVECTOR) return false; if (riscv_autovec_preference == RVV_SCALABLE) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 91af2c00674..46f1a1da33e 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -115,23 +115,37 @@ public: tree type = builtin_types[e.type.index].vector; machine_mode mode = TYPE_MODE (type); - /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. - - - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 - - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 - - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 - - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 - - e8,m2/e16,m4/e32,m8 --> e8m2 - - e8,m4/e16,m8 --> e8m4 - - e8,m8 --> e8m8 - */ - /* SEW. */ - e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); - - /* LMUL. */ - machine_mode e8_mode - = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); - e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + + if (TARGET_XTHEADVECTOR) + { + machine_mode inner_mode = GET_MODE_INNER (mode); + /* SEW. */ + e.add_input_operand (Pmode, + gen_int_mode (GET_MODE_BITSIZE (inner_mode), Pmode)); + /* LMUL. */ + e.add_input_operand (Pmode, + gen_int_mode (get_vlmul (mode), Pmode)); + } + else + { + /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. + + - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 + - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 + - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 + - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 + - e8,m2/e16,m4/e32,m8 --> e8m2 + - e8,m4/e16,m8 --> e8m4 + - e8,m8 --> e8m8 + */ + /* SEW. */ + e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); + + /* LMUL. */ + machine_mode e8_mode + = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); + e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + } /* TAIL_ANY. */ e.add_input_operand (Pmode, diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index ee8058dcdb1..1e4f4d53de6 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -33,6 +33,25 @@ namespace riscv_vector { +/* Check whether the RETURN_TYPE and ARGUMENT_TYPES are + valid for the function. */ + +static bool +check_type (tree return_type, vec<tree> &argument_types) +{ + tree arg; + unsigned i; + + if (!return_type) + return false; + + FOR_EACH_VEC_ELT (argument_types, i, arg) + if (!arg) + return false; + + return true; +} + /* Add one function instance for GROUP, using operand suffix at index OI, mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ static void @@ -49,6 +68,10 @@ build_one (function_builder &b, const function_group_info &group, group.ops_infos.types[vec_type_idx].index); b.allocate_argument_types (function_instance, argument_types); b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + b.add_overloaded_function (function_instance, *group.shape); b.add_unique_function (function_instance, (*group.shape), return_type, argument_types); diff --git a/gcc/config/riscv/riscv-vector-switch.def b/gcc/config/riscv/riscv-vector-switch.def index 1ad26c2a3b2..452283b7416 100644 --- a/gcc/config/riscv/riscv-vector-switch.def +++ b/gcc/config/riscv/riscv-vector-switch.def @@ -68,9 +68,9 @@ Encode the ratio of SEW/LMUL into the mask types. #endif /* Disable modes if TARGET_MIN_VLEN == 32. */ -ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) -ENTRY (RVVMF32BI, true, LMUL_F4, 32) -ENTRY (RVVMF16BI, true, LMUL_F2, 16) +ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F8, 64) +ENTRY (RVVMF32BI, true, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F4, 32) +ENTRY (RVVMF16BI, true, TARGET_XTHEADVECTOR ? LMUL_1 : LMUL_F2 , 16) ENTRY (RVVMF8BI, true, LMUL_1, 8) ENTRY (RVVMF4BI, true, LMUL_2, 4) ENTRY (RVVMF2BI, true, LMUL_4, 2) @@ -81,39 +81,39 @@ ENTRY (RVVM8QI, true, LMUL_8, 1) ENTRY (RVVM4QI, true, LMUL_4, 2) ENTRY (RVVM2QI, true, LMUL_2, 4) ENTRY (RVVM1QI, true, LMUL_1, 8) -ENTRY (RVVMF2QI, true, LMUL_F2, 16) -ENTRY (RVVMF4QI, true, LMUL_F4, 32) -ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) +ENTRY (RVVMF2QI, !TARGET_XTHEADVECTOR, LMUL_F2, 16) +ENTRY (RVVMF4QI, !TARGET_XTHEADVECTOR, LMUL_F4, 32) +ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F8, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8HI, true, LMUL_8, 2) ENTRY (RVVM4HI, true, LMUL_4, 4) ENTRY (RVVM2HI, true, LMUL_2, 8) ENTRY (RVVM1HI, true, LMUL_1, 16) -ENTRY (RVVMF2HI, true, LMUL_F2, 32) -ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HI, !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_16. */ ENTRY (RVVM8HF, TARGET_VECTOR_ELEN_FP_16, LMUL_8, 2) ENTRY (RVVM4HF, TARGET_VECTOR_ELEN_FP_16, LMUL_4, 4) ENTRY (RVVM2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_2, 8) ENTRY (RVVM1HF, TARGET_VECTOR_ELEN_FP_16, LMUL_1, 16) -ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_F2, 32) -ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8SI, true, LMUL_8, 4) ENTRY (RVVM4SI, true, LMUL_4, 8) ENTRY (RVVM2SI, true, LMUL_2, 16) ENTRY (RVVM1SI, true, LMUL_1, 32) -ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_32. */ ENTRY (RVVM8SF, TARGET_VECTOR_ELEN_FP_32, LMUL_8, 4) ENTRY (RVVM4SF, TARGET_VECTOR_ELEN_FP_32, LMUL_4, 8) ENTRY (RVVM2SF, TARGET_VECTOR_ELEN_FP_32, LMUL_2, 16) ENTRY (RVVM1SF, TARGET_VECTOR_ELEN_FP_32, LMUL_1, 32) -ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if !TARGET_VECTOR_ELEN_64. */ ENTRY (RVVM8DI, TARGET_VECTOR_ELEN_64, LMUL_8, 8) @@ -140,127 +140,127 @@ ENTRY (RVVM1DF, TARGET_VECTOR_ELEN_FP_64, LMUL_1, 64) #endif TUPLE_ENTRY (RVVM1x8QI, true, RVVM1QI, 8, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x8QI, true, RVVMF2QI, 8, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x8QI, true, RVVMF4QI, 8, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 8, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x8QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 8, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x8QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 8, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 8, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x7QI, true, RVVM1QI, 7, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x7QI, true, RVVMF2QI, 7, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x7QI, true, RVVMF4QI, 7, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 7, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x7QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 7, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x7QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 7, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 7, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x6QI, true, RVVM1QI, 6, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x6QI, true, RVVMF2QI, 6, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x6QI, true, RVVMF4QI, 6, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 6, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x6QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 6, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x6QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 6, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 6, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x5QI, true, RVVM1QI, 5, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x5QI, true, RVVMF2QI, 5, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x5QI, true, RVVMF4QI, 5, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 5, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x5QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 5, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x5QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 5, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 5, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x4QI, true, RVVM2QI, 4, LMUL_2, 4) TUPLE_ENTRY (RVVM1x4QI, true, RVVM1QI, 4, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x4QI, true, RVVMF2QI, 4, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x4QI, true, RVVMF4QI, 4, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 4, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x4QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 4, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x4QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 4, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 4, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x3QI, true, RVVM2QI, 3, LMUL_2, 4) TUPLE_ENTRY (RVVM1x3QI, true, RVVM1QI, 3, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x3QI, true, RVVMF2QI, 3, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x3QI, true, RVVMF4QI, 3, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 3, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x3QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 3, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x3QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 3, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 3, LMUL_F8, 64) TUPLE_ENTRY (RVVM4x2QI, true, RVVM4QI, 2, LMUL_4, 2) TUPLE_ENTRY (RVVM2x2QI, true, RVVM2QI, 2, LMUL_2, 4) TUPLE_ENTRY (RVVM1x2QI, true, RVVM1QI, 2, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x2QI, true, RVVMF2QI, 2, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x2QI, true, RVVMF4QI, 2, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 2, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x2QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 2, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x2QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 2, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 2, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x8HI, true, RVVM1HI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HI, true, RVVMF2HI, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HI, true, RVVM1HI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HI, true, RVVMF2HI, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HI, true, RVVM1HI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HI, true, RVVMF2HI, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HI, true, RVVM1HI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HI, true, RVVMF2HI, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HI, true, RVVM2HI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HI, true, RVVM1HI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HI, true, RVVMF2HI, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HI, true, RVVM2HI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HI, true, RVVM1HI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HI, true, RVVMF2HI, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HI, true, RVVM4HI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HI, true, RVVM2HI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HI, true, RVVM1HI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HI, true, RVVMF2HI, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM4HF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8SI, true, RVVM1SI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SI, true, RVVM1SI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SI, true, RVVM1SI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SI, true, RVVM1SI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SI, true, RVVM2SI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SI, true, RVVM1SI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SI, true, RVVM2SI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SI, true, RVVM1SI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SI, true, RVVM4SI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SI, true, RVVM2SI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SI, true, RVVM1SI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM4SF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 8, LMUL_1, 16) TUPLE_ENTRY (RVVM1x7DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 7, LMUL_1, 16) diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index 8f74cd2e502..28cf1f414e9 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -1406,6 +1406,9 @@ riscv_v_adjust_bytesize (machine_mode mode, int scale) { if (riscv_v_ext_vector_mode_p (mode)) { + if (TARGET_XTHEADVECTOR) + return BYTES_PER_RISCV_VECTOR; + poly_int64 nunits = GET_MODE_NUNITS (mode); poly_int64 mode_size = GET_MODE_SIZE (mode); @@ -9995,7 +9998,7 @@ riscv_use_divmod_expander (void) static machine_mode riscv_preferred_simd_mode (scalar_mode mode) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::preferred_simd_mode (mode); return word_mode; @@ -10346,7 +10349,7 @@ riscv_mode_priority (int, int n) unsigned int riscv_autovectorize_vector_modes (vector_modes *modes, bool all) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::autovectorize_vector_modes (modes, all); return default_autovectorize_vector_modes (modes, all); @@ -10529,6 +10532,16 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) return false; } +/* Implements target hook vector_mode_supported_any_target_p. */ + +static bool +riscv_vector_mode_supported_any_target_p (machine_mode mode) +{ + if (TARGET_XTHEADVECTOR) + return false; + return true; +} + /* Initialize the GCC target structure. */ #undef TARGET_ASM_ALIGNED_HI_OP #define TARGET_ASM_ALIGNED_HI_OP "\t.half\t" @@ -10872,6 +10885,9 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) #undef TARGET_PREFERRED_ELSE_VALUE #define TARGET_PREFERRED_ELSE_VALUE riscv_preferred_else_value +#undef TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P +#define TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P riscv_vector_mode_supported_any_target_p + struct gcc_target targetm = TARGET_INITIALIZER; #include "gt-riscv.h" diff --git a/gcc/config/riscv/riscv_th_vector.h b/gcc/config/riscv/riscv_th_vector.h new file mode 100644 index 00000000000..b6b6738bdda --- /dev/null +++ b/gcc/config/riscv/riscv_th_vector.h @@ -0,0 +1,49 @@ +/* RISC-V 'XTheadVector' Extension intrinsics include file. + Copyright (C) 2024 Free Software Foundation, Inc. + + This file is part of GCC. + + GCC is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published + by the Free Software Foundation; either version 3, or (at your + option) any later version. + + GCC is distributed in the hope that it will be useful, but WITHOUT + ANY WARRANTY; without even the implied warranty of MERCHANTABILITY + or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public + License for more details. + + Under Section 7 of GPL version 3, you are granted additional + permissions described in the GCC Runtime Library Exception, version + 3.1, as published by the Free Software Foundation. + + You should have received a copy of the GNU General Public License and + a copy of the GCC Runtime Library Exception along with this program; + see the files COPYING3 and COPYING.RUNTIME respectively. If not, see + <" target="_blank">http://www.gnu.org/licenses/>. */ + +#ifndef __RISCV_TH_VECTOR_H +#define __RISCV_TH_VECTOR_H + +#include <stdint.h> +#include <stddef.h> + +#ifndef __riscv_xtheadvector +#error "XTheadVector intrinsics require the xtheadvector extension." +#else +#ifdef __cplusplus +extern "C" { +#endif + +/* NOTE: This implementation of riscv_th_vector.h is intentionally short. It does + not define the RVV types and intrinsic functions directly in C and C++ + code, but instead uses the following pragma to tell GCC to insert the + necessary type and function definitions itself. The net effect is the + same, and the file is a complete implementation of riscv_th_vector.h. */ +#pragma riscv intrinsic "xtheadvector" + +#ifdef __cplusplus +} +#endif // __cplusplus +#endif // __riscv_xtheadvector +#endif // __RISCV_TH_ECTOR_H diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md new file mode 100644 index 00000000000..696b815252d --- /dev/null +++ b/gcc/config/riscv/thead-vector.md @@ -0,0 +1,102 @@ +(define_c_enum "unspec" [ + UNSPEC_TH_VWLDST +]) + +(define_mode_iterator V_VLS_VT [V VLS VT]) +(define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) + +(define_split + [(set (match_operand:V_VB_VLS_VT 0 "reg_or_mem_operand") + (match_operand:V_VB_VLS_VT 1 "reg_or_mem_operand"))] + "TARGET_XTHEADVECTOR" + [(const_int 0)] + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + }) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:V_VLS_VT 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:V_VLS_VT + [(match_operand:V_VLS_VT 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2")]) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:VB 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:VB + [(match_operand:VB 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2") + (set (attr "sew") (const_int 8)) + (set (attr "vlmul") (symbol_ref "riscv_vector::LMUL_1"))]) + +(define_insn_and_split "*pred_th_mov<mode>" + [(set (match_operand:VB_VLS 0 "nonimmediate_operand" "=vr, m, vr, vr, vr") + (if_then_else:VB_VLS + (unspec:VB_VLS + [(match_operand:VB_VLS 1 "vector_all_trues_mask_operand" "Wc1, Wc1, Wc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) + (match_operand:VB_VLS 3 "vector_move_operand" " m, vr, vr, Wc0, Wc1") + (match_operand:VB_VLS 2 "vector_undef_operand" " vu, vu, vu, vu, vu")))] + "TARGET_XTHEADVECTOR" + "@ + # + # + vmcpy.m\t%0,%3 + vmclr.m\t%0 + vmset.m\t%0" + "&& !reload_completed" + [(const_int 0)] + { + if ((MEM_P (operands[0]) || MEM_P (operands[3])) + || (REG_P (operands[0]) && REG_P (operands[3]) + && INTVAL (operands[5]) == riscv_vector::VLMAX)) + { + emit_move_insn (operands[0], operands[3]); + DONE; + } + FAIL; + } + [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/config/riscv/thead.cc b/gcc/config/riscv/thead.cc index 00b2e14c6c0..cc04101b256 100644 --- a/gcc/config/riscv/thead.cc +++ b/gcc/config/riscv/thead.cc @@ -27,6 +27,7 @@ #include "backend.h" #include "tree.h" #include "rtl.h" +#include "insn-attr.h" #include "explow.h" #include "memmodel.h" #include "emit-rtl.h" @@ -890,8 +891,26 @@ th_asm_output_opcode (FILE *asm_out_file, const char *p) { /* We need to add th. prefix to all the xtheadvector instructions here.*/ - if (current_output_insn != NULL && p[0] == 'v') - fputs ("th.", asm_out_file); + if (current_output_insn != NULL) + { + if (get_attr_type (current_output_insn) == TYPE_VSETVL) + { + if (strstr (p, "zero")) + { + if (strstr (p, "zero,zero")) + return "th.vsetvli\tzero,zero,e%0,%m1"; + else + return "th.vsetvli\tzero,%0,e%1,%m2"; + } + else + { + return "th.vsetvli\t%0,%1,e%2,%m3"; + } + } + + if (p[0] == 'v') + fputs ("th.", asm_out_file); + } return p; } diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index 24b7b4394be..3eb6daafbc2 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -87,7 +87,7 @@ ;; check. However, we need default value of SEW for vsetvl instruction since there ;; is no field for ratio in the vsetvl instruction encoding. (define_attr "sew" "" - (cond [(eq_attr "mode" "RVVMF64BI,RVVMF32BI,RVVMF16BI,RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ + (cond [(eq_attr "mode" "RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ RVVM8QI,RVVM4QI,RVVM2QI,RVVM1QI,RVVMF2QI,RVVMF4QI,RVVMF8QI,\ RVVM1x8QI,RVVMF2x8QI,RVVMF4x8QI,RVVMF8x8QI,\ RVVM1x7QI,RVVMF2x7QI,RVVMF4x7QI,RVVMF8x7QI,\ @@ -99,6 +99,18 @@ V1QI,V2QI,V4QI,V8QI,V16QI,V32QI,V64QI,V128QI,V256QI,V512QI,V1024QI,V2048QI,V4096QI,\ V1BI,V2BI,V4BI,V8BI,V16BI,V32BI,V64BI,V128BI,V256BI,V512BI,V1024BI,V2048BI,V4096BI") (const_int 8) + (eq_attr "mode" "RVVMF16BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 16) + (const_int 8)) + (eq_attr "mode" "RVVMF32BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 32) + (const_int 8)) + (eq_attr "mode" "RVVMF64BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 64) + (const_int 8)) (eq_attr "mode" "RVVM8HI,RVVM4HI,RVVM2HI,RVVM1HI,RVVMF2HI,RVVMF4HI,\ RVVM1x8HI,RVVMF2x8HI,RVVMF4x8HI,\ RVVM1x7HI,RVVMF2x7HI,RVVMF4x7HI,\ @@ -159,9 +171,9 @@ (eq_attr "mode" "RVVM4QI,RVVMF2BI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2QI,RVVMF4BI") (symbol_ref "riscv_vector::LMUL_2") (eq_attr "mode" "RVVM1QI,RVVMF8BI") (symbol_ref "riscv_vector::LMUL_1") - (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "riscv_vector::LMUL_F2") - (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "riscv_vector::LMUL_F4") - (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "riscv_vector::LMUL_F8") + (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F2") + (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F4") + (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F8") (eq_attr "mode" "RVVM8HI") (symbol_ref "riscv_vector::LMUL_8") (eq_attr "mode" "RVVM4HI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2HI") (symbol_ref "riscv_vector::LMUL_2") @@ -436,6 +448,10 @@ vaeskf1,vaeskf2,vaesz,vsha2ms,vsha2ch,vsha2cl,vsm4k,vsm4r,\ vsm3me,vsm3c") (const_int INVALID_ATTRIBUTE) + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE) (eq_attr "mode" "RVVM8QI,RVVM1BI") (const_int 1) (eq_attr "mode" "RVVM4QI,RVVMF2BI") (const_int 2) (eq_attr "mode" "RVVM2QI,RVVMF4BI") (const_int 4) @@ -905,6 +921,8 @@ (symbol_ref "riscv_vector::FRM_DYN")] (symbol_ref "riscv_vector::FRM_NONE"))) +(include "thead-vector.md") + ;; ----------------------------------------------------------------- ;; ---- Miscellaneous Operations ;; ----------------------------------------------------------------- @@ -1078,6 +1096,12 @@ - We can not leave it to TARGET_SECONDARY_RELOAD since it happens before spilling. The clobber scratch is used by spilling fractional registers in IRA/LRA so it's too early. */ + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; @@ -1114,7 +1138,7 @@ (define_insn "*mov<mode>_whole" [(set (match_operand:V_WHOLE 0 "reg_or_mem_operand" "=vr, m,vr") (match_operand:V_WHOLE 1 "reg_or_mem_operand" " m,vr,vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "@ vl%m1re<sew>.v\t%0,%1 vs%m1r.v\t%1,%0 @@ -1135,6 +1159,13 @@ (match_operand:VB 1 "general_operand"))] "TARGET_VECTOR" { + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } + if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; }) @@ -1142,7 +1173,7 @@ (define_insn "*mov<mode>" [(set (match_operand:VB 0 "register_operand" "=vr") (match_operand:VB 1 "register_operand" " vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "vmv1r.v\t%0,%1" [(set_attr "type" "vmov") (set_attr "mode" "<MODE>")]) @@ -3692,7 +3723,7 @@ (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3713,7 +3744,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3734,7 +3765,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c index 2e0e12aa045..2eef9e1e1a8 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c @@ -1,4 +1,4 @@ -/* { dg-do compile } */ +/* { dg-do compile { target { ! riscv_xtheadvector } } } */ /* { dg-skip-if "test rvv intrinsic" { *-*-* } { "*" } { "-march=rv*v*" } } */ void foo0 () {__rvv_bool64_t t;} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c index 3d81b179235..ef329e30785 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c @@ -1,4 +1,4 @@ /* { dg-do compile } */ /* { dg-options "-O3 -march=rv32gc -mabi=ilp32d" } */ -#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' extension enabled} } */ +#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' or 'XTHEADVECTOR' extension enabled} } */ diff --git a/gcc/testsuite/lib/target-supports.exp b/gcc/testsuite/lib/target-supports.exp index b27c30b8c51..bf0ad5342d6 100644 --- a/gcc/testsuite/lib/target-supports.exp +++ b/gcc/testsuite/lib/target-supports.exp @@ -1956,6 +1956,18 @@ proc check_effective_target_riscv_zbb { } { }] } +# Return 1 if the target arch supports the XTheadVector extension, 0 otherwise. +# Cache the result. + +proc check_effective_target_riscv_xtheadvector { } { + return [check_no_compiler_messages riscv_ext_xtheadvector assembly { + #ifndef __riscv_xtheadvector + #error "Not __riscv_xtheadvector" + #endif + }] +} + + # Return 1 if we can execute code when using dg-add-options riscv_v proc check_effective_target_riscv_v_ok { } { -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re：Re：[PATCH v5] RISC-V: Handle differences between XTheadvector and Vector 2024-01-10 7:01 ` juzhe.zhong @ 2024-01-10 7:16 ` joshua 2024-01-10 7:17 ` Re：[PATCH " juzhe.zhong 0 siblings, 1 reply; 133+ messages in thread From: joshua @ 2024-01-10 7:16 UTC (permalink / raw) To: juzhe.zhong, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu These xttheadvector speical intrinsics are different from rvv1.0 in determining function name from base name. We cannot directly reuse the existing shapes. In order not to invade existing shapes, we add new shapes for new functions. Also, we create new thead-vector-builtins.cc for xtheadvector function_base implementation. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月10日(星期三) 15:01 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Handle differences between XTheadvector and Vector Why do you add theadvector shapes ? I think you can reuse the current existing shapes. +thead-vector-builtins.o: \+ $(srcdir)/config/riscv/thead-vector-builtins.cc \+ $(CONFIG_H) $(SYSTEM_H) coretypes.h $(TM_H) $(TREE_H) $(RTL_H) \+ $(TM_P_H) memmodel.h insn-codes.h $(OPTABS_H) $(RECOG_H) \+ $(EXPR_H) $(BASIC_BLOCK_H) $(FUNCTION_H) fold-const.h $(GIMPLE_H) \+ gimple-iterator.h gimplify.h explow.h $(EMIT_RTL_H) tree-vector-builder.h \+ rtx-vector-builder.h \+ $(srcdir)/config/riscv/riscv-vector-builtins-shapes.h \+ $(srcdir)/config/riscv/riscv-vector-builtins-bases.h \+ $(srcdir)/config/riscv/thead-vector-builtins.h \+ $(RISCV_BUILTINS_H)+ $(COMPILER) -c $(ALL_COMPILERFLAGS) $(ALL_CPPFLAGS) $(INCLUDES) \+ $(srcdir)/config/riscv/thead-vector-builtins.cc+ Why do you rebuild another new object ? + Copyright (C) 2022-2023 Free Software Foundation, Inc. Incorrect copyright juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-10 10:57 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Handle differences between XTheadvector and Vector Hi Juzhe, Thank you for so many useful comments for this patch! There are some more patches to support xtheadvector special instrinsics as well as handle register overlap issue and rewrite assembly output. https://gcc.gnu.org/pipermail/gcc-patches/2024-January/641774.html https://gcc.gnu.org/pipermail/gcc-patches/2024-January/641732.html https://gcc.gnu.org/pipermail/gcc-patches/2024-January/641733.html Also, there is a precedent patch to refactor riscv-vector-builtins-bases.cc https://gcc.gnu.org/pipermail/gcc-patches/2023-December/641530.html Jeff has reviewed it, but didn't have LGTM yet. Joshua ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月10日(星期三) 10:34 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Handle differences between XTheadvector and Vector Thanks for your patience. LGTM from myside. I think it's pretty clean now. I can image in the future when some day the theadvector is no longer used, we can remove it very easily. And also, the theadvector won't affect our RVV1.0 maintain since it's isolated cleanly. But I'd like to wait for a few more days some body want to chime in. And you should do more things before commit it: 1. Remember you should run the full coverage RVV1.0 API test, the test-generator is downloaded from official intrinsic doc: https://github.com/riscv-non-isa/rvv-intrinsic-doc 2. Also the regression of RV32 an RV64 of GCC testsuite. Do you have more patches of theadvector that I didn't review ? plz point them to me again. Thanks. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-10 10:22 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Handle differences between XTheadvector and Vector This patch is to handle the differences in instruction generation between Vector and XTheadVector. In this version, we only support partial xtheadvector instructions that leverage directly from current RVV1.0 with simple adding "th." prefix. For different name xtheadvector instructions but share same patterns as RVV1.0 instructions, we will use ASM targethook to rewrite the whole string of the instructions in the following patches. For some vector patterns that cannot be avoided, we use "!TARGET_XTHEADVECTOR" to disable them in vector.md in order not to generate instructions that xtheadvector does not support, like vmv1r and vsext.vf2. gcc/ChangeLog: * config.gcc: Add files for XTheadVector intrinsics. * config/riscv/autovec.md: Guard XTheadVector. * config/riscv/riscv-c.cc: Add pragma for XTheadVector. * config/riscv/riscv-string.cc (expand_block_move): Guard XTheadVector. * config/riscv/riscv-string.cc (vls_mode_valid_p): Avoid autovec. * config/riscv/riscv-vector-builtins-shapes.cc (check_type): (build_one): New function. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): (DEF_THEAD_RVV_FUNCTION): Add new marcos. (check_required_extensions): (handle_pragma_vector): * config/riscv/riscv-vector-builtins.h (RVV_REQUIRE_VECTOR): (RVV_REQUIRE_XTHEADVECTOR): Add RVV_REQUIRE_VECTOR and RVV_REQUIRE_XTHEADVECTOR. (struct function_group_info): * config/riscv/riscv-vector-switch.def (ENTRY): Disable fractional mode for the XTheadVector extension. (TUPLE_ENTRY): Likewise. * config/riscv/riscv-vsetvl.cc: Add functions for xtheadvector. * config/riscv/riscv.cc (riscv_v_ext_vls_mode_p): Guard XTheadVector. (riscv_v_adjust_bytesize): Likewise. (riscv_preferred_simd_mode): Likewsie. (riscv_autovectorize_vector_modes): Likewise. (riscv_vector_mode_supported_any_target_p): Likewise. (TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P): Likewise. * config/riscv/vector.md: Include thead-vector.md. * config/riscv/riscv_th_vector.h: New file. * config/riscv/thead-vector.md: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/base/pragma-1.c: Add XTheadVector. * gcc.target/riscv/rvv/base/abi-1.c: Exclude XTheadVector. * lib/target-supports.exp: Add target for XTheadVector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config.gcc | 2 +- gcc/config/riscv/autovec.md | 2 +- gcc/config/riscv/predicates.md | 4 +- gcc/config/riscv/riscv-c.cc | 3 +- gcc/config/riscv/riscv-string.cc | 3 +- gcc/config/riscv/riscv-v.cc | 2 +- .../riscv/riscv-vector-builtins-bases.cc | 48 ++++-- .../riscv/riscv-vector-builtins-shapes.cc | 23 +++ gcc/config/riscv/riscv-vector-switch.def | 150 +++++++++--------- gcc/config/riscv/riscv.cc | 20 ++- gcc/config/riscv/riscv_th_vector.h | 49 ++++++ gcc/config/riscv/thead-vector.md | 102 ++++++++++++ gcc/config/riscv/thead.cc | 23 ++- gcc/config/riscv/vector.md | 49 ++++-- .../gcc.target/riscv/rvv/base/abi-1.c | 2 +- .../gcc.target/riscv/rvv/base/pragma-1.c | 2 +- gcc/testsuite/lib/target-supports.exp | 12 ++ 17 files changed, 383 insertions(+), 113 deletions(-) create mode 100644 gcc/config/riscv/riscv_th_vector.h create mode 100644 gcc/config/riscv/thead-vector.md diff --git a/gcc/config.gcc b/gcc/config.gcc index 7e583390024..047e4c02cf4 100644 --- a/gcc/config.gcc +++ b/gcc/config.gcc @@ -549,7 +549,7 @@ riscv*) extra_objs="${extra_objs} riscv-vector-builtins.o riscv-vector-builtins-shapes.o riscv-vector-builtins-bases.o" extra_objs="${extra_objs} thead.o riscv-target-attr.o" d_target_objs="riscv-d.o" - extra_headers="riscv_vector.h" + extra_headers="riscv_vector.h riscv_th_vector.h" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.cc" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.h" ;; diff --git a/gcc/config/riscv/autovec.md b/gcc/config/riscv/autovec.md index 775eaa825b0..0477781cabe 100644 --- a/gcc/config/riscv/autovec.md +++ b/gcc/config/riscv/autovec.md @@ -2579,7 +2579,7 @@ [(match_operand 0 "register_operand") (match_operand 1 "memory_operand") (match_operand:ANYI 2 "const_int_operand")] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" { riscv_vector::expand_rawmemchr(<MODE>mode, operands[0], operands[1], operands[2]); diff --git a/gcc/config/riscv/predicates.md b/gcc/config/riscv/predicates.md index b1a79cae50a..0337da88284 100644 --- a/gcc/config/riscv/predicates.md +++ b/gcc/config/riscv/predicates.md @@ -428,7 +428,9 @@ ;; Predicates for the V extension. (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (ior (match_test "TARGET_XTHEADVECTOR && rtx_equal_p (op, const0_rtx)") + (match_test "!TARGET_XTHEADVECTOR")) + (match_operand 0 "const_csr_operand")))) (define_special_predicate "autovec_length_operand" (ior (match_operand 0 "pmode_register_operand") diff --git a/gcc/config/riscv/riscv-c.cc b/gcc/config/riscv/riscv-c.cc index 422ddc2c308..2e306057347 100644 --- a/gcc/config/riscv/riscv-c.cc +++ b/gcc/config/riscv/riscv-c.cc @@ -195,7 +195,8 @@ riscv_pragma_intrinsic (cpp_reader *) const char *name = TREE_STRING_POINTER (x); - if (strcmp (name, "vector") == 0) + if (strcmp (name, "vector") == 0 + || strcmp (name, "xtheadvector") == 0) { if (!TARGET_VECTOR) { diff --git a/gcc/config/riscv/riscv-string.cc b/gcc/config/riscv/riscv-string.cc index f227b8667ce..b09b51d7526 100644 --- a/gcc/config/riscv/riscv-string.cc +++ b/gcc/config/riscv/riscv-string.cc @@ -773,7 +773,8 @@ riscv_expand_block_move_scalar (rtx dest, rtx src, rtx length) bool riscv_expand_block_move (rtx dest, rtx src, rtx length) { - if (TARGET_VECTOR && stringop_strategy & STRATEGY_VECTOR) + if ((TARGET_VECTOR && !TARGET_XTHEADVECTOR) + && stringop_strategy & STRATEGY_VECTOR) { bool ok = riscv_vector::expand_block_move (dest, src, length); if (ok) diff --git a/gcc/config/riscv/riscv-v.cc b/gcc/config/riscv/riscv-v.cc index 2491522191a..7e25deb966d 100644 --- a/gcc/config/riscv/riscv-v.cc +++ b/gcc/config/riscv/riscv-v.cc @@ -4392,7 +4392,7 @@ cmp_lmul_gt_one (machine_mode mode) bool vls_mode_valid_p (machine_mode vls_mode) { - if (!TARGET_VECTOR) + if (!TARGET_VECTOR || TARGET_XTHEADVECTOR) return false; if (riscv_autovec_preference == RVV_SCALABLE) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 91af2c00674..46f1a1da33e 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -115,23 +115,37 @@ public: tree type = builtin_types[e.type.index].vector; machine_mode mode = TYPE_MODE (type); - /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. - - - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 - - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 - - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 - - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 - - e8,m2/e16,m4/e32,m8 --> e8m2 - - e8,m4/e16,m8 --> e8m4 - - e8,m8 --> e8m8 - */ - /* SEW. */ - e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); - - /* LMUL. */ - machine_mode e8_mode - = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); - e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + + if (TARGET_XTHEADVECTOR) + { + machine_mode inner_mode = GET_MODE_INNER (mode); + /* SEW. */ + e.add_input_operand (Pmode, + gen_int_mode (GET_MODE_BITSIZE (inner_mode), Pmode)); + /* LMUL. */ + e.add_input_operand (Pmode, + gen_int_mode (get_vlmul (mode), Pmode)); + } + else + { + /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. + + - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 + - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 + - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 + - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 + - e8,m2/e16,m4/e32,m8 --> e8m2 + - e8,m4/e16,m8 --> e8m4 + - e8,m8 --> e8m8 + */ + /* SEW. */ + e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); + + /* LMUL. */ + machine_mode e8_mode + = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); + e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + } /* TAIL_ANY. */ e.add_input_operand (Pmode, diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index ee8058dcdb1..1e4f4d53de6 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -33,6 +33,25 @@ namespace riscv_vector { +/* Check whether the RETURN_TYPE and ARGUMENT_TYPES are + valid for the function. */ + +static bool +check_type (tree return_type, vec<tree> &argument_types) +{ + tree arg; + unsigned i; + + if (!return_type) + return false; + + FOR_EACH_VEC_ELT (argument_types, i, arg) + if (!arg) + return false; + + return true; +} + /* Add one function instance for GROUP, using operand suffix at index OI, mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ static void @@ -49,6 +68,10 @@ build_one (function_builder &b, const function_group_info &group, group.ops_infos.types[vec_type_idx].index); b.allocate_argument_types (function_instance, argument_types); b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + b.add_overloaded_function (function_instance, *group.shape); b.add_unique_function (function_instance, (*group.shape), return_type, argument_types); diff --git a/gcc/config/riscv/riscv-vector-switch.def b/gcc/config/riscv/riscv-vector-switch.def index 1ad26c2a3b2..452283b7416 100644 --- a/gcc/config/riscv/riscv-vector-switch.def +++ b/gcc/config/riscv/riscv-vector-switch.def @@ -68,9 +68,9 @@ Encode the ratio of SEW/LMUL into the mask types. #endif /* Disable modes if TARGET_MIN_VLEN == 32. */ -ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) -ENTRY (RVVMF32BI, true, LMUL_F4, 32) -ENTRY (RVVMF16BI, true, LMUL_F2, 16) +ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F8, 64) +ENTRY (RVVMF32BI, true, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F4, 32) +ENTRY (RVVMF16BI, true, TARGET_XTHEADVECTOR ? LMUL_1 : LMUL_F2 , 16) ENTRY (RVVMF8BI, true, LMUL_1, 8) ENTRY (RVVMF4BI, true, LMUL_2, 4) ENTRY (RVVMF2BI, true, LMUL_4, 2) @@ -81,39 +81,39 @@ ENTRY (RVVM8QI, true, LMUL_8, 1) ENTRY (RVVM4QI, true, LMUL_4, 2) ENTRY (RVVM2QI, true, LMUL_2, 4) ENTRY (RVVM1QI, true, LMUL_1, 8) -ENTRY (RVVMF2QI, true, LMUL_F2, 16) -ENTRY (RVVMF4QI, true, LMUL_F4, 32) -ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) +ENTRY (RVVMF2QI, !TARGET_XTHEADVECTOR, LMUL_F2, 16) +ENTRY (RVVMF4QI, !TARGET_XTHEADVECTOR, LMUL_F4, 32) +ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F8, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8HI, true, LMUL_8, 2) ENTRY (RVVM4HI, true, LMUL_4, 4) ENTRY (RVVM2HI, true, LMUL_2, 8) ENTRY (RVVM1HI, true, LMUL_1, 16) -ENTRY (RVVMF2HI, true, LMUL_F2, 32) -ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HI, !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_16. */ ENTRY (RVVM8HF, TARGET_VECTOR_ELEN_FP_16, LMUL_8, 2) ENTRY (RVVM4HF, TARGET_VECTOR_ELEN_FP_16, LMUL_4, 4) ENTRY (RVVM2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_2, 8) ENTRY (RVVM1HF, TARGET_VECTOR_ELEN_FP_16, LMUL_1, 16) -ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_F2, 32) -ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8SI, true, LMUL_8, 4) ENTRY (RVVM4SI, true, LMUL_4, 8) ENTRY (RVVM2SI, true, LMUL_2, 16) ENTRY (RVVM1SI, true, LMUL_1, 32) -ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_32. */ ENTRY (RVVM8SF, TARGET_VECTOR_ELEN_FP_32, LMUL_8, 4) ENTRY (RVVM4SF, TARGET_VECTOR_ELEN_FP_32, LMUL_4, 8) ENTRY (RVVM2SF, TARGET_VECTOR_ELEN_FP_32, LMUL_2, 16) ENTRY (RVVM1SF, TARGET_VECTOR_ELEN_FP_32, LMUL_1, 32) -ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if !TARGET_VECTOR_ELEN_64. */ ENTRY (RVVM8DI, TARGET_VECTOR_ELEN_64, LMUL_8, 8) @@ -140,127 +140,127 @@ ENTRY (RVVM1DF, TARGET_VECTOR_ELEN_FP_64, LMUL_1, 64) #endif TUPLE_ENTRY (RVVM1x8QI, true, RVVM1QI, 8, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x8QI, true, RVVMF2QI, 8, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x8QI, true, RVVMF4QI, 8, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 8, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x8QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 8, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x8QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 8, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 8, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x7QI, true, RVVM1QI, 7, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x7QI, true, RVVMF2QI, 7, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x7QI, true, RVVMF4QI, 7, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 7, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x7QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 7, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x7QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 7, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 7, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x6QI, true, RVVM1QI, 6, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x6QI, true, RVVMF2QI, 6, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x6QI, true, RVVMF4QI, 6, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 6, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x6QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 6, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x6QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 6, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 6, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x5QI, true, RVVM1QI, 5, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x5QI, true, RVVMF2QI, 5, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x5QI, true, RVVMF4QI, 5, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 5, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x5QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 5, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x5QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 5, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 5, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x4QI, true, RVVM2QI, 4, LMUL_2, 4) TUPLE_ENTRY (RVVM1x4QI, true, RVVM1QI, 4, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x4QI, true, RVVMF2QI, 4, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x4QI, true, RVVMF4QI, 4, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 4, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x4QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 4, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x4QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 4, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 4, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x3QI, true, RVVM2QI, 3, LMUL_2, 4) TUPLE_ENTRY (RVVM1x3QI, true, RVVM1QI, 3, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x3QI, true, RVVMF2QI, 3, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x3QI, true, RVVMF4QI, 3, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 3, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x3QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 3, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x3QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 3, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 3, LMUL_F8, 64) TUPLE_ENTRY (RVVM4x2QI, true, RVVM4QI, 2, LMUL_4, 2) TUPLE_ENTRY (RVVM2x2QI, true, RVVM2QI, 2, LMUL_2, 4) TUPLE_ENTRY (RVVM1x2QI, true, RVVM1QI, 2, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x2QI, true, RVVMF2QI, 2, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x2QI, true, RVVMF4QI, 2, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 2, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x2QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 2, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x2QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 2, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 2, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x8HI, true, RVVM1HI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HI, true, RVVMF2HI, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HI, true, RVVM1HI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HI, true, RVVMF2HI, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HI, true, RVVM1HI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HI, true, RVVMF2HI, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HI, true, RVVM1HI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HI, true, RVVMF2HI, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HI, true, RVVM2HI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HI, true, RVVM1HI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HI, true, RVVMF2HI, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HI, true, RVVM2HI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HI, true, RVVM1HI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HI, true, RVVMF2HI, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HI, true, RVVM4HI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HI, true, RVVM2HI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HI, true, RVVM1HI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HI, true, RVVMF2HI, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM4HF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8SI, true, RVVM1SI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SI, true, RVVM1SI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SI, true, RVVM1SI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SI, true, RVVM1SI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SI, true, RVVM2SI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SI, true, RVVM1SI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SI, true, RVVM2SI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SI, true, RVVM1SI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SI, true, RVVM4SI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SI, true, RVVM2SI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SI, true, RVVM1SI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM4SF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 8, LMUL_1, 16) TUPLE_ENTRY (RVVM1x7DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 7, LMUL_1, 16) diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index 8f74cd2e502..28cf1f414e9 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -1406,6 +1406,9 @@ riscv_v_adjust_bytesize (machine_mode mode, int scale) { if (riscv_v_ext_vector_mode_p (mode)) { + if (TARGET_XTHEADVECTOR) + return BYTES_PER_RISCV_VECTOR; + poly_int64 nunits = GET_MODE_NUNITS (mode); poly_int64 mode_size = GET_MODE_SIZE (mode); @@ -9995,7 +9998,7 @@ riscv_use_divmod_expander (void) static machine_mode riscv_preferred_simd_mode (scalar_mode mode) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::preferred_simd_mode (mode); return word_mode; @@ -10346,7 +10349,7 @@ riscv_mode_priority (int, int n) unsigned int riscv_autovectorize_vector_modes (vector_modes *modes, bool all) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::autovectorize_vector_modes (modes, all); return default_autovectorize_vector_modes (modes, all); @@ -10529,6 +10532,16 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) return false; } +/* Implements target hook vector_mode_supported_any_target_p. */ + +static bool +riscv_vector_mode_supported_any_target_p (machine_mode mode) +{ + if (TARGET_XTHEADVECTOR) + return false; + return true; +} + /* Initialize the GCC target structure. */ #undef TARGET_ASM_ALIGNED_HI_OP #define TARGET_ASM_ALIGNED_HI_OP "\t.half\t" @@ -10872,6 +10885,9 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) #undef TARGET_PREFERRED_ELSE_VALUE #define TARGET_PREFERRED_ELSE_VALUE riscv_preferred_else_value +#undef TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P +#define TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P riscv_vector_mode_supported_any_target_p + struct gcc_target targetm = TARGET_INITIALIZER; #include "gt-riscv.h" diff --git a/gcc/config/riscv/riscv_th_vector.h b/gcc/config/riscv/riscv_th_vector.h new file mode 100644 index 00000000000..b6b6738bdda --- /dev/null +++ b/gcc/config/riscv/riscv_th_vector.h @@ -0,0 +1,49 @@ +/* RISC-V 'XTheadVector' Extension intrinsics include file. + Copyright (C) 2024 Free Software Foundation, Inc. + + This file is part of GCC. + + GCC is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published + by the Free Software Foundation; either version 3, or (at your + option) any later version. + + GCC is distributed in the hope that it will be useful, but WITHOUT + ANY WARRANTY; without even the implied warranty of MERCHANTABILITY + or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public + License for more details. + + Under Section 7 of GPL version 3, you are granted additional + permissions described in the GCC Runtime Library Exception, version + 3.1, as published by the Free Software Foundation. + + You should have received a copy of the GNU General Public License and + a copy of the GCC Runtime Library Exception along with this program; + see the files COPYING3 and COPYING.RUNTIME respectively. If not, see + <" target="_blank">" target="_blank">http://www.gnu.org/licenses/>. */ + +#ifndef __RISCV_TH_VECTOR_H +#define __RISCV_TH_VECTOR_H + +#include <stdint.h> +#include <stddef.h> + +#ifndef __riscv_xtheadvector +#error "XTheadVector intrinsics require the xtheadvector extension." +#else +#ifdef __cplusplus +extern "C" { +#endif + +/* NOTE: This implementation of riscv_th_vector.h is intentionally short. It does + not define the RVV types and intrinsic functions directly in C and C++ + code, but instead uses the following pragma to tell GCC to insert the + necessary type and function definitions itself. The net effect is the + same, and the file is a complete implementation of riscv_th_vector.h. */ +#pragma riscv intrinsic "xtheadvector" + +#ifdef __cplusplus +} +#endif // __cplusplus +#endif // __riscv_xtheadvector +#endif // __RISCV_TH_ECTOR_H diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md new file mode 100644 index 00000000000..696b815252d --- /dev/null +++ b/gcc/config/riscv/thead-vector.md @@ -0,0 +1,102 @@ +(define_c_enum "unspec" [ + UNSPEC_TH_VWLDST +]) + +(define_mode_iterator V_VLS_VT [V VLS VT]) +(define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) + +(define_split + [(set (match_operand:V_VB_VLS_VT 0 "reg_or_mem_operand") + (match_operand:V_VB_VLS_VT 1 "reg_or_mem_operand"))] + "TARGET_XTHEADVECTOR" + [(const_int 0)] + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + }) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:V_VLS_VT 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:V_VLS_VT + [(match_operand:V_VLS_VT 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2")]) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:VB 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:VB + [(match_operand:VB 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2") + (set (attr "sew") (const_int 8)) + (set (attr "vlmul") (symbol_ref "riscv_vector::LMUL_1"))]) + +(define_insn_and_split "*pred_th_mov<mode>" + [(set (match_operand:VB_VLS 0 "nonimmediate_operand" "=vr, m, vr, vr, vr") + (if_then_else:VB_VLS + (unspec:VB_VLS + [(match_operand:VB_VLS 1 "vector_all_trues_mask_operand" "Wc1, Wc1, Wc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) + (match_operand:VB_VLS 3 "vector_move_operand" " m, vr, vr, Wc0, Wc1") + (match_operand:VB_VLS 2 "vector_undef_operand" " vu, vu, vu, vu, vu")))] + "TARGET_XTHEADVECTOR" + "@ + # + # + vmcpy.m\t%0,%3 + vmclr.m\t%0 + vmset.m\t%0" + "&& !reload_completed" + [(const_int 0)] + { + if ((MEM_P (operands[0]) || MEM_P (operands[3])) + || (REG_P (operands[0]) && REG_P (operands[3]) + && INTVAL (operands[5]) == riscv_vector::VLMAX)) + { + emit_move_insn (operands[0], operands[3]); + DONE; + } + FAIL; + } + [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/config/riscv/thead.cc b/gcc/config/riscv/thead.cc index 00b2e14c6c0..cc04101b256 100644 --- a/gcc/config/riscv/thead.cc +++ b/gcc/config/riscv/thead.cc @@ -27,6 +27,7 @@ #include "backend.h" #include "tree.h" #include "rtl.h" +#include "insn-attr.h" #include "explow.h" #include "memmodel.h" #include "emit-rtl.h" @@ -890,8 +891,26 @@ th_asm_output_opcode (FILE *asm_out_file, const char *p) { /* We need to add th. prefix to all the xtheadvector instructions here.*/ - if (current_output_insn != NULL && p[0] == 'v') - fputs ("th.", asm_out_file); + if (current_output_insn != NULL) + { + if (get_attr_type (current_output_insn) == TYPE_VSETVL) + { + if (strstr (p, "zero")) + { + if (strstr (p, "zero,zero")) + return "th.vsetvli\tzero,zero,e%0,%m1"; + else + return "th.vsetvli\tzero,%0,e%1,%m2"; + } + else + { + return "th.vsetvli\t%0,%1,e%2,%m3"; + } + } + + if (p[0] == 'v') + fputs ("th.", asm_out_file); + } return p; } diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index 24b7b4394be..3eb6daafbc2 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -87,7 +87,7 @@ ;; check. However, we need default value of SEW for vsetvl instruction since there ;; is no field for ratio in the vsetvl instruction encoding. (define_attr "sew" "" - (cond [(eq_attr "mode" "RVVMF64BI,RVVMF32BI,RVVMF16BI,RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ + (cond [(eq_attr "mode" "RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ RVVM8QI,RVVM4QI,RVVM2QI,RVVM1QI,RVVMF2QI,RVVMF4QI,RVVMF8QI,\ RVVM1x8QI,RVVMF2x8QI,RVVMF4x8QI,RVVMF8x8QI,\ RVVM1x7QI,RVVMF2x7QI,RVVMF4x7QI,RVVMF8x7QI,\ @@ -99,6 +99,18 @@ V1QI,V2QI,V4QI,V8QI,V16QI,V32QI,V64QI,V128QI,V256QI,V512QI,V1024QI,V2048QI,V4096QI,\ V1BI,V2BI,V4BI,V8BI,V16BI,V32BI,V64BI,V128BI,V256BI,V512BI,V1024BI,V2048BI,V4096BI") (const_int 8) + (eq_attr "mode" "RVVMF16BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 16) + (const_int 8)) + (eq_attr "mode" "RVVMF32BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 32) + (const_int 8)) + (eq_attr "mode" "RVVMF64BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 64) + (const_int 8)) (eq_attr "mode" "RVVM8HI,RVVM4HI,RVVM2HI,RVVM1HI,RVVMF2HI,RVVMF4HI,\ RVVM1x8HI,RVVMF2x8HI,RVVMF4x8HI,\ RVVM1x7HI,RVVMF2x7HI,RVVMF4x7HI,\ @@ -159,9 +171,9 @@ (eq_attr "mode" "RVVM4QI,RVVMF2BI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2QI,RVVMF4BI") (symbol_ref "riscv_vector::LMUL_2") (eq_attr "mode" "RVVM1QI,RVVMF8BI") (symbol_ref "riscv_vector::LMUL_1") - (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "riscv_vector::LMUL_F2") - (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "riscv_vector::LMUL_F4") - (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "riscv_vector::LMUL_F8") + (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F2") + (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F4") + (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F8") (eq_attr "mode" "RVVM8HI") (symbol_ref "riscv_vector::LMUL_8") (eq_attr "mode" "RVVM4HI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2HI") (symbol_ref "riscv_vector::LMUL_2") @@ -436,6 +448,10 @@ vaeskf1,vaeskf2,vaesz,vsha2ms,vsha2ch,vsha2cl,vsm4k,vsm4r,\ vsm3me,vsm3c") (const_int INVALID_ATTRIBUTE) + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE) (eq_attr "mode" "RVVM8QI,RVVM1BI") (const_int 1) (eq_attr "mode" "RVVM4QI,RVVMF2BI") (const_int 2) (eq_attr "mode" "RVVM2QI,RVVMF4BI") (const_int 4) @@ -905,6 +921,8 @@ (symbol_ref "riscv_vector::FRM_DYN")] (symbol_ref "riscv_vector::FRM_NONE"))) +(include "thead-vector.md") + ;; ----------------------------------------------------------------- ;; ---- Miscellaneous Operations ;; ----------------------------------------------------------------- @@ -1078,6 +1096,12 @@ - We can not leave it to TARGET_SECONDARY_RELOAD since it happens before spilling. The clobber scratch is used by spilling fractional registers in IRA/LRA so it's too early. */ + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; @@ -1114,7 +1138,7 @@ (define_insn "*mov<mode>_whole" [(set (match_operand:V_WHOLE 0 "reg_or_mem_operand" "=vr, m,vr") (match_operand:V_WHOLE 1 "reg_or_mem_operand" " m,vr,vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "@ vl%m1re<sew>.v\t%0,%1 vs%m1r.v\t%1,%0 @@ -1135,6 +1159,13 @@ (match_operand:VB 1 "general_operand"))] "TARGET_VECTOR" { + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } + if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; }) @@ -1142,7 +1173,7 @@ (define_insn "*mov<mode>" [(set (match_operand:VB 0 "register_operand" "=vr") (match_operand:VB 1 "register_operand" " vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "vmv1r.v\t%0,%1" [(set_attr "type" "vmov") (set_attr "mode" "<MODE>")]) @@ -3692,7 +3723,7 @@ (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3713,7 +3744,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3734,7 +3765,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c index 2e0e12aa045..2eef9e1e1a8 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c @@ -1,4 +1,4 @@ -/* { dg-do compile } */ +/* { dg-do compile { target { ! riscv_xtheadvector } } } */ /* { dg-skip-if "test rvv intrinsic" { *-*-* } { "*" } { "-march=rv*v*" } } */ void foo0 () {__rvv_bool64_t t;} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c index 3d81b179235..ef329e30785 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c @@ -1,4 +1,4 @@ /* { dg-do compile } */ /* { dg-options "-O3 -march=rv32gc -mabi=ilp32d" } */ -#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' extension enabled} } */ +#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' or 'XTHEADVECTOR' extension enabled} } */ diff --git a/gcc/testsuite/lib/target-supports.exp b/gcc/testsuite/lib/target-supports.exp index b27c30b8c51..bf0ad5342d6 100644 --- a/gcc/testsuite/lib/target-supports.exp +++ b/gcc/testsuite/lib/target-supports.exp @@ -1956,6 +1956,18 @@ proc check_effective_target_riscv_zbb { } { }] } +# Return 1 if the target arch supports the XTheadVector extension, 0 otherwise. +# Cache the result. + +proc check_effective_target_riscv_xtheadvector { } { + return [check_no_compiler_messages riscv_ext_xtheadvector assembly { + #ifndef __riscv_xtheadvector + #error "Not __riscv_xtheadvector" + #endif + }] +} + + # Return 1 if we can execute code when using dg-add-options riscv_v proc check_effective_target_riscv_v_ok { } { -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: Re：[PATCH v5] RISC-V: Handle differences between XTheadvector and Vector 2024-01-10 7:16 ` Re：Re：[PATCH " joshua @ 2024-01-10 7:17 ` juzhe.zhong 2024-01-10 7:26 ` 回复：Re：[PATCH " joshua 2024-01-10 7:28 ` Re：Re：[PATCH " joshua 0 siblings, 2 replies; 133+ messages in thread From: juzhe.zhong @ 2024-01-10 7:17 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 50296 bytes --] Why do you need to invade existing shapes ? juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-10 15:16 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Handle differences between XTheadvector and Vector These xttheadvector speical intrinsics are different from rvv1.0 in determining function name from base name. We cannot directly reuse the existing shapes. In order not to invade existing shapes, we add new shapes for new functions. Also, we create new thead-vector-builtins.cc for xtheadvector function_base implementation. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月10日(星期三) 15:01 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Handle differences between XTheadvector and Vector Why do you add theadvector shapes ? I think you can reuse the current existing shapes. +thead-vector-builtins.o: \+ $(srcdir)/config/riscv/thead-vector-builtins.cc \+ $(CONFIG_H) $(SYSTEM_H) coretypes.h $(TM_H) $(TREE_H) $(RTL_H) \+ $(TM_P_H) memmodel.h insn-codes.h $(OPTABS_H) $(RECOG_H) \+ $(EXPR_H) $(BASIC_BLOCK_H) $(FUNCTION_H) fold-const.h $(GIMPLE_H) \+ gimple-iterator.h gimplify.h explow.h $(EMIT_RTL_H) tree-vector-builder.h \+ rtx-vector-builder.h \+ $(srcdir)/config/riscv/riscv-vector-builtins-shapes.h \+ $(srcdir)/config/riscv/riscv-vector-builtins-bases.h \+ $(srcdir)/config/riscv/thead-vector-builtins.h \+ $(RISCV_BUILTINS_H)+ $(COMPILER) -c $(ALL_COMPILERFLAGS) $(ALL_CPPFLAGS) $(INCLUDES) \+ $(srcdir)/config/riscv/thead-vector-builtins.cc+ Why do you rebuild another new object ? + Copyright (C) 2022-2023 Free Software Foundation, Inc. Incorrect copyright juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-10 10:57 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Handle differences between XTheadvector and Vector Hi Juzhe, Thank you for so many useful comments for this patch! There are some more patches to support xtheadvector special instrinsics as well as handle register overlap issue and rewrite assembly output. https://gcc.gnu.org/pipermail/gcc-patches/2024-January/641774.html https://gcc.gnu.org/pipermail/gcc-patches/2024-January/641732.html https://gcc.gnu.org/pipermail/gcc-patches/2024-January/641733.html Also, there is a precedent patch to refactor riscv-vector-builtins-bases.cc https://gcc.gnu.org/pipermail/gcc-patches/2023-December/641530.html Jeff has reviewed it, but didn't have LGTM yet. Joshua ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月10日(星期三) 10:34 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Handle differences between XTheadvector and Vector Thanks for your patience. LGTM from myside. I think it's pretty clean now. I can image in the future when some day the theadvector is no longer used, we can remove it very easily. And also, the theadvector won't affect our RVV1.0 maintain since it's isolated cleanly. But I'd like to wait for a few more days some body want to chime in. And you should do more things before commit it: 1. Remember you should run the full coverage RVV1.0 API test, the test-generator is downloaded from official intrinsic doc: https://github.com/riscv-non-isa/rvv-intrinsic-doc 2. Also the regression of RV32 an RV64 of GCC testsuite. Do you have more patches of theadvector that I didn't review ? plz point them to me again. Thanks. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-10 10:22 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Handle differences between XTheadvector and Vector This patch is to handle the differences in instruction generation between Vector and XTheadVector. In this version, we only support partial xtheadvector instructions that leverage directly from current RVV1.0 with simple adding "th." prefix. For different name xtheadvector instructions but share same patterns as RVV1.0 instructions, we will use ASM targethook to rewrite the whole string of the instructions in the following patches. For some vector patterns that cannot be avoided, we use "!TARGET_XTHEADVECTOR" to disable them in vector.md in order not to generate instructions that xtheadvector does not support, like vmv1r and vsext.vf2. gcc/ChangeLog: * config.gcc: Add files for XTheadVector intrinsics. * config/riscv/autovec.md: Guard XTheadVector. * config/riscv/riscv-c.cc: Add pragma for XTheadVector. * config/riscv/riscv-string.cc (expand_block_move): Guard XTheadVector. * config/riscv/riscv-string.cc (vls_mode_valid_p): Avoid autovec. * config/riscv/riscv-vector-builtins-shapes.cc (check_type): (build_one): New function. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): (DEF_THEAD_RVV_FUNCTION): Add new marcos. (check_required_extensions): (handle_pragma_vector): * config/riscv/riscv-vector-builtins.h (RVV_REQUIRE_VECTOR): (RVV_REQUIRE_XTHEADVECTOR): Add RVV_REQUIRE_VECTOR and RVV_REQUIRE_XTHEADVECTOR. (struct function_group_info): * config/riscv/riscv-vector-switch.def (ENTRY): Disable fractional mode for the XTheadVector extension. (TUPLE_ENTRY): Likewise. * config/riscv/riscv-vsetvl.cc: Add functions for xtheadvector. * config/riscv/riscv.cc (riscv_v_ext_vls_mode_p): Guard XTheadVector. (riscv_v_adjust_bytesize): Likewise. (riscv_preferred_simd_mode): Likewsie. (riscv_autovectorize_vector_modes): Likewise. (riscv_vector_mode_supported_any_target_p): Likewise. (TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P): Likewise. * config/riscv/vector.md: Include thead-vector.md. * config/riscv/riscv_th_vector.h: New file. * config/riscv/thead-vector.md: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/base/pragma-1.c: Add XTheadVector. * gcc.target/riscv/rvv/base/abi-1.c: Exclude XTheadVector. * lib/target-supports.exp: Add target for XTheadVector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config.gcc | 2 +- gcc/config/riscv/autovec.md | 2 +- gcc/config/riscv/predicates.md | 4 +- gcc/config/riscv/riscv-c.cc | 3 +- gcc/config/riscv/riscv-string.cc | 3 +- gcc/config/riscv/riscv-v.cc | 2 +- .../riscv/riscv-vector-builtins-bases.cc | 48 ++++-- .../riscv/riscv-vector-builtins-shapes.cc | 23 +++ gcc/config/riscv/riscv-vector-switch.def | 150 +++++++++--------- gcc/config/riscv/riscv.cc | 20 ++- gcc/config/riscv/riscv_th_vector.h | 49 ++++++ gcc/config/riscv/thead-vector.md | 102 ++++++++++++ gcc/config/riscv/thead.cc | 23 ++- gcc/config/riscv/vector.md | 49 ++++-- .../gcc.target/riscv/rvv/base/abi-1.c | 2 +- .../gcc.target/riscv/rvv/base/pragma-1.c | 2 +- gcc/testsuite/lib/target-supports.exp | 12 ++ 17 files changed, 383 insertions(+), 113 deletions(-) create mode 100644 gcc/config/riscv/riscv_th_vector.h create mode 100644 gcc/config/riscv/thead-vector.md diff --git a/gcc/config.gcc b/gcc/config.gcc index 7e583390024..047e4c02cf4 100644 --- a/gcc/config.gcc +++ b/gcc/config.gcc @@ -549,7 +549,7 @@ riscv*) extra_objs="${extra_objs} riscv-vector-builtins.o riscv-vector-builtins-shapes.o riscv-vector-builtins-bases.o" extra_objs="${extra_objs} thead.o riscv-target-attr.o" d_target_objs="riscv-d.o" - extra_headers="riscv_vector.h" + extra_headers="riscv_vector.h riscv_th_vector.h" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.cc" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.h" ;; diff --git a/gcc/config/riscv/autovec.md b/gcc/config/riscv/autovec.md index 775eaa825b0..0477781cabe 100644 --- a/gcc/config/riscv/autovec.md +++ b/gcc/config/riscv/autovec.md @@ -2579,7 +2579,7 @@ [(match_operand 0 "register_operand") (match_operand 1 "memory_operand") (match_operand:ANYI 2 "const_int_operand")] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" { riscv_vector::expand_rawmemchr(<MODE>mode, operands[0], operands[1], operands[2]); diff --git a/gcc/config/riscv/predicates.md b/gcc/config/riscv/predicates.md index b1a79cae50a..0337da88284 100644 --- a/gcc/config/riscv/predicates.md +++ b/gcc/config/riscv/predicates.md @@ -428,7 +428,9 @@ ;; Predicates for the V extension. (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (ior (match_test "TARGET_XTHEADVECTOR && rtx_equal_p (op, const0_rtx)") + (match_test "!TARGET_XTHEADVECTOR")) + (match_operand 0 "const_csr_operand")))) (define_special_predicate "autovec_length_operand" (ior (match_operand 0 "pmode_register_operand") diff --git a/gcc/config/riscv/riscv-c.cc b/gcc/config/riscv/riscv-c.cc index 422ddc2c308..2e306057347 100644 --- a/gcc/config/riscv/riscv-c.cc +++ b/gcc/config/riscv/riscv-c.cc @@ -195,7 +195,8 @@ riscv_pragma_intrinsic (cpp_reader *) const char *name = TREE_STRING_POINTER (x); - if (strcmp (name, "vector") == 0) + if (strcmp (name, "vector") == 0 + || strcmp (name, "xtheadvector") == 0) { if (!TARGET_VECTOR) { diff --git a/gcc/config/riscv/riscv-string.cc b/gcc/config/riscv/riscv-string.cc index f227b8667ce..b09b51d7526 100644 --- a/gcc/config/riscv/riscv-string.cc +++ b/gcc/config/riscv/riscv-string.cc @@ -773,7 +773,8 @@ riscv_expand_block_move_scalar (rtx dest, rtx src, rtx length) bool riscv_expand_block_move (rtx dest, rtx src, rtx length) { - if (TARGET_VECTOR && stringop_strategy & STRATEGY_VECTOR) + if ((TARGET_VECTOR && !TARGET_XTHEADVECTOR) + && stringop_strategy & STRATEGY_VECTOR) { bool ok = riscv_vector::expand_block_move (dest, src, length); if (ok) diff --git a/gcc/config/riscv/riscv-v.cc b/gcc/config/riscv/riscv-v.cc index 2491522191a..7e25deb966d 100644 --- a/gcc/config/riscv/riscv-v.cc +++ b/gcc/config/riscv/riscv-v.cc @@ -4392,7 +4392,7 @@ cmp_lmul_gt_one (machine_mode mode) bool vls_mode_valid_p (machine_mode vls_mode) { - if (!TARGET_VECTOR) + if (!TARGET_VECTOR || TARGET_XTHEADVECTOR) return false; if (riscv_autovec_preference == RVV_SCALABLE) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 91af2c00674..46f1a1da33e 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -115,23 +115,37 @@ public: tree type = builtin_types[e.type.index].vector; machine_mode mode = TYPE_MODE (type); - /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. - - - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 - - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 - - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 - - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 - - e8,m2/e16,m4/e32,m8 --> e8m2 - - e8,m4/e16,m8 --> e8m4 - - e8,m8 --> e8m8 - */ - /* SEW. */ - e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); - - /* LMUL. */ - machine_mode e8_mode - = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); - e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + + if (TARGET_XTHEADVECTOR) + { + machine_mode inner_mode = GET_MODE_INNER (mode); + /* SEW. */ + e.add_input_operand (Pmode, + gen_int_mode (GET_MODE_BITSIZE (inner_mode), Pmode)); + /* LMUL. */ + e.add_input_operand (Pmode, + gen_int_mode (get_vlmul (mode), Pmode)); + } + else + { + /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. + + - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 + - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 + - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 + - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 + - e8,m2/e16,m4/e32,m8 --> e8m2 + - e8,m4/e16,m8 --> e8m4 + - e8,m8 --> e8m8 + */ + /* SEW. */ + e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); + + /* LMUL. */ + machine_mode e8_mode + = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); + e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + } /* TAIL_ANY. */ e.add_input_operand (Pmode, diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index ee8058dcdb1..1e4f4d53de6 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -33,6 +33,25 @@ namespace riscv_vector { +/* Check whether the RETURN_TYPE and ARGUMENT_TYPES are + valid for the function. */ + +static bool +check_type (tree return_type, vec<tree> &argument_types) +{ + tree arg; + unsigned i; + + if (!return_type) + return false; + + FOR_EACH_VEC_ELT (argument_types, i, arg) + if (!arg) + return false; + + return true; +} + /* Add one function instance for GROUP, using operand suffix at index OI, mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ static void @@ -49,6 +68,10 @@ build_one (function_builder &b, const function_group_info &group, group.ops_infos.types[vec_type_idx].index); b.allocate_argument_types (function_instance, argument_types); b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + b.add_overloaded_function (function_instance, *group.shape); b.add_unique_function (function_instance, (*group.shape), return_type, argument_types); diff --git a/gcc/config/riscv/riscv-vector-switch.def b/gcc/config/riscv/riscv-vector-switch.def index 1ad26c2a3b2..452283b7416 100644 --- a/gcc/config/riscv/riscv-vector-switch.def +++ b/gcc/config/riscv/riscv-vector-switch.def @@ -68,9 +68,9 @@ Encode the ratio of SEW/LMUL into the mask types. #endif /* Disable modes if TARGET_MIN_VLEN == 32. */ -ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) -ENTRY (RVVMF32BI, true, LMUL_F4, 32) -ENTRY (RVVMF16BI, true, LMUL_F2, 16) +ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F8, 64) +ENTRY (RVVMF32BI, true, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F4, 32) +ENTRY (RVVMF16BI, true, TARGET_XTHEADVECTOR ? LMUL_1 : LMUL_F2 , 16) ENTRY (RVVMF8BI, true, LMUL_1, 8) ENTRY (RVVMF4BI, true, LMUL_2, 4) ENTRY (RVVMF2BI, true, LMUL_4, 2) @@ -81,39 +81,39 @@ ENTRY (RVVM8QI, true, LMUL_8, 1) ENTRY (RVVM4QI, true, LMUL_4, 2) ENTRY (RVVM2QI, true, LMUL_2, 4) ENTRY (RVVM1QI, true, LMUL_1, 8) -ENTRY (RVVMF2QI, true, LMUL_F2, 16) -ENTRY (RVVMF4QI, true, LMUL_F4, 32) -ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) +ENTRY (RVVMF2QI, !TARGET_XTHEADVECTOR, LMUL_F2, 16) +ENTRY (RVVMF4QI, !TARGET_XTHEADVECTOR, LMUL_F4, 32) +ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F8, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8HI, true, LMUL_8, 2) ENTRY (RVVM4HI, true, LMUL_4, 4) ENTRY (RVVM2HI, true, LMUL_2, 8) ENTRY (RVVM1HI, true, LMUL_1, 16) -ENTRY (RVVMF2HI, true, LMUL_F2, 32) -ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HI, !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_16. */ ENTRY (RVVM8HF, TARGET_VECTOR_ELEN_FP_16, LMUL_8, 2) ENTRY (RVVM4HF, TARGET_VECTOR_ELEN_FP_16, LMUL_4, 4) ENTRY (RVVM2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_2, 8) ENTRY (RVVM1HF, TARGET_VECTOR_ELEN_FP_16, LMUL_1, 16) -ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_F2, 32) -ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8SI, true, LMUL_8, 4) ENTRY (RVVM4SI, true, LMUL_4, 8) ENTRY (RVVM2SI, true, LMUL_2, 16) ENTRY (RVVM1SI, true, LMUL_1, 32) -ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_32. */ ENTRY (RVVM8SF, TARGET_VECTOR_ELEN_FP_32, LMUL_8, 4) ENTRY (RVVM4SF, TARGET_VECTOR_ELEN_FP_32, LMUL_4, 8) ENTRY (RVVM2SF, TARGET_VECTOR_ELEN_FP_32, LMUL_2, 16) ENTRY (RVVM1SF, TARGET_VECTOR_ELEN_FP_32, LMUL_1, 32) -ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if !TARGET_VECTOR_ELEN_64. */ ENTRY (RVVM8DI, TARGET_VECTOR_ELEN_64, LMUL_8, 8) @@ -140,127 +140,127 @@ ENTRY (RVVM1DF, TARGET_VECTOR_ELEN_FP_64, LMUL_1, 64) #endif TUPLE_ENTRY (RVVM1x8QI, true, RVVM1QI, 8, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x8QI, true, RVVMF2QI, 8, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x8QI, true, RVVMF4QI, 8, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 8, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x8QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 8, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x8QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 8, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 8, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x7QI, true, RVVM1QI, 7, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x7QI, true, RVVMF2QI, 7, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x7QI, true, RVVMF4QI, 7, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 7, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x7QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 7, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x7QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 7, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 7, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x6QI, true, RVVM1QI, 6, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x6QI, true, RVVMF2QI, 6, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x6QI, true, RVVMF4QI, 6, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 6, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x6QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 6, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x6QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 6, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 6, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x5QI, true, RVVM1QI, 5, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x5QI, true, RVVMF2QI, 5, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x5QI, true, RVVMF4QI, 5, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 5, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x5QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 5, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x5QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 5, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 5, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x4QI, true, RVVM2QI, 4, LMUL_2, 4) TUPLE_ENTRY (RVVM1x4QI, true, RVVM1QI, 4, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x4QI, true, RVVMF2QI, 4, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x4QI, true, RVVMF4QI, 4, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 4, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x4QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 4, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x4QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 4, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 4, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x3QI, true, RVVM2QI, 3, LMUL_2, 4) TUPLE_ENTRY (RVVM1x3QI, true, RVVM1QI, 3, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x3QI, true, RVVMF2QI, 3, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x3QI, true, RVVMF4QI, 3, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 3, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x3QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 3, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x3QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 3, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 3, LMUL_F8, 64) TUPLE_ENTRY (RVVM4x2QI, true, RVVM4QI, 2, LMUL_4, 2) TUPLE_ENTRY (RVVM2x2QI, true, RVVM2QI, 2, LMUL_2, 4) TUPLE_ENTRY (RVVM1x2QI, true, RVVM1QI, 2, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x2QI, true, RVVMF2QI, 2, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x2QI, true, RVVMF4QI, 2, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 2, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x2QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 2, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x2QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 2, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 2, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x8HI, true, RVVM1HI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HI, true, RVVMF2HI, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HI, true, RVVM1HI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HI, true, RVVMF2HI, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HI, true, RVVM1HI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HI, true, RVVMF2HI, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HI, true, RVVM1HI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HI, true, RVVMF2HI, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HI, true, RVVM2HI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HI, true, RVVM1HI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HI, true, RVVMF2HI, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HI, true, RVVM2HI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HI, true, RVVM1HI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HI, true, RVVMF2HI, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HI, true, RVVM4HI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HI, true, RVVM2HI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HI, true, RVVM1HI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HI, true, RVVMF2HI, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM4HF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8SI, true, RVVM1SI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SI, true, RVVM1SI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SI, true, RVVM1SI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SI, true, RVVM1SI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SI, true, RVVM2SI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SI, true, RVVM1SI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SI, true, RVVM2SI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SI, true, RVVM1SI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SI, true, RVVM4SI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SI, true, RVVM2SI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SI, true, RVVM1SI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM4SF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 8, LMUL_1, 16) TUPLE_ENTRY (RVVM1x7DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 7, LMUL_1, 16) diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index 8f74cd2e502..28cf1f414e9 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -1406,6 +1406,9 @@ riscv_v_adjust_bytesize (machine_mode mode, int scale) { if (riscv_v_ext_vector_mode_p (mode)) { + if (TARGET_XTHEADVECTOR) + return BYTES_PER_RISCV_VECTOR; + poly_int64 nunits = GET_MODE_NUNITS (mode); poly_int64 mode_size = GET_MODE_SIZE (mode); @@ -9995,7 +9998,7 @@ riscv_use_divmod_expander (void) static machine_mode riscv_preferred_simd_mode (scalar_mode mode) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::preferred_simd_mode (mode); return word_mode; @@ -10346,7 +10349,7 @@ riscv_mode_priority (int, int n) unsigned int riscv_autovectorize_vector_modes (vector_modes *modes, bool all) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::autovectorize_vector_modes (modes, all); return default_autovectorize_vector_modes (modes, all); @@ -10529,6 +10532,16 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) return false; } +/* Implements target hook vector_mode_supported_any_target_p. */ + +static bool +riscv_vector_mode_supported_any_target_p (machine_mode mode) +{ + if (TARGET_XTHEADVECTOR) + return false; + return true; +} + /* Initialize the GCC target structure. */ #undef TARGET_ASM_ALIGNED_HI_OP #define TARGET_ASM_ALIGNED_HI_OP "\t.half\t" @@ -10872,6 +10885,9 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) #undef TARGET_PREFERRED_ELSE_VALUE #define TARGET_PREFERRED_ELSE_VALUE riscv_preferred_else_value +#undef TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P +#define TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P riscv_vector_mode_supported_any_target_p + struct gcc_target targetm = TARGET_INITIALIZER; #include "gt-riscv.h" diff --git a/gcc/config/riscv/riscv_th_vector.h b/gcc/config/riscv/riscv_th_vector.h new file mode 100644 index 00000000000..b6b6738bdda --- /dev/null +++ b/gcc/config/riscv/riscv_th_vector.h @@ -0,0 +1,49 @@ +/* RISC-V 'XTheadVector' Extension intrinsics include file. + Copyright (C) 2024 Free Software Foundation, Inc. + + This file is part of GCC. + + GCC is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published + by the Free Software Foundation; either version 3, or (at your + option) any later version. + + GCC is distributed in the hope that it will be useful, but WITHOUT + ANY WARRANTY; without even the implied warranty of MERCHANTABILITY + or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public + License for more details. + + Under Section 7 of GPL version 3, you are granted additional + permissions described in the GCC Runtime Library Exception, version + 3.1, as published by the Free Software Foundation. + + You should have received a copy of the GNU General Public License and + a copy of the GCC Runtime Library Exception along with this program; + see the files COPYING3 and COPYING.RUNTIME respectively. If not, see + <" target="_blank">" target="_blank">http://www.gnu.org/licenses/>. */ + +#ifndef __RISCV_TH_VECTOR_H +#define __RISCV_TH_VECTOR_H + +#include <stdint.h> +#include <stddef.h> + +#ifndef __riscv_xtheadvector +#error "XTheadVector intrinsics require the xtheadvector extension." +#else +#ifdef __cplusplus +extern "C" { +#endif + +/* NOTE: This implementation of riscv_th_vector.h is intentionally short. It does + not define the RVV types and intrinsic functions directly in C and C++ + code, but instead uses the following pragma to tell GCC to insert the + necessary type and function definitions itself. The net effect is the + same, and the file is a complete implementation of riscv_th_vector.h. */ +#pragma riscv intrinsic "xtheadvector" + +#ifdef __cplusplus +} +#endif // __cplusplus +#endif // __riscv_xtheadvector +#endif // __RISCV_TH_ECTOR_H diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md new file mode 100644 index 00000000000..696b815252d --- /dev/null +++ b/gcc/config/riscv/thead-vector.md @@ -0,0 +1,102 @@ +(define_c_enum "unspec" [ + UNSPEC_TH_VWLDST +]) + +(define_mode_iterator V_VLS_VT [V VLS VT]) +(define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) + +(define_split + [(set (match_operand:V_VB_VLS_VT 0 "reg_or_mem_operand") + (match_operand:V_VB_VLS_VT 1 "reg_or_mem_operand"))] + "TARGET_XTHEADVECTOR" + [(const_int 0)] + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + }) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:V_VLS_VT 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:V_VLS_VT + [(match_operand:V_VLS_VT 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2")]) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:VB 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:VB + [(match_operand:VB 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2") + (set (attr "sew") (const_int 8)) + (set (attr "vlmul") (symbol_ref "riscv_vector::LMUL_1"))]) + +(define_insn_and_split "*pred_th_mov<mode>" + [(set (match_operand:VB_VLS 0 "nonimmediate_operand" "=vr, m, vr, vr, vr") + (if_then_else:VB_VLS + (unspec:VB_VLS + [(match_operand:VB_VLS 1 "vector_all_trues_mask_operand" "Wc1, Wc1, Wc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) + (match_operand:VB_VLS 3 "vector_move_operand" " m, vr, vr, Wc0, Wc1") + (match_operand:VB_VLS 2 "vector_undef_operand" " vu, vu, vu, vu, vu")))] + "TARGET_XTHEADVECTOR" + "@ + # + # + vmcpy.m\t%0,%3 + vmclr.m\t%0 + vmset.m\t%0" + "&& !reload_completed" + [(const_int 0)] + { + if ((MEM_P (operands[0]) || MEM_P (operands[3])) + || (REG_P (operands[0]) && REG_P (operands[3]) + && INTVAL (operands[5]) == riscv_vector::VLMAX)) + { + emit_move_insn (operands[0], operands[3]); + DONE; + } + FAIL; + } + [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/config/riscv/thead.cc b/gcc/config/riscv/thead.cc index 00b2e14c6c0..cc04101b256 100644 --- a/gcc/config/riscv/thead.cc +++ b/gcc/config/riscv/thead.cc @@ -27,6 +27,7 @@ #include "backend.h" #include "tree.h" #include "rtl.h" +#include "insn-attr.h" #include "explow.h" #include "memmodel.h" #include "emit-rtl.h" @@ -890,8 +891,26 @@ th_asm_output_opcode (FILE *asm_out_file, const char *p) { /* We need to add th. prefix to all the xtheadvector instructions here.*/ - if (current_output_insn != NULL && p[0] == 'v') - fputs ("th.", asm_out_file); + if (current_output_insn != NULL) + { + if (get_attr_type (current_output_insn) == TYPE_VSETVL) + { + if (strstr (p, "zero")) + { + if (strstr (p, "zero,zero")) + return "th.vsetvli\tzero,zero,e%0,%m1"; + else + return "th.vsetvli\tzero,%0,e%1,%m2"; + } + else + { + return "th.vsetvli\t%0,%1,e%2,%m3"; + } + } + + if (p[0] == 'v') + fputs ("th.", asm_out_file); + } return p; } diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index 24b7b4394be..3eb6daafbc2 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -87,7 +87,7 @@ ;; check. However, we need default value of SEW for vsetvl instruction since there ;; is no field for ratio in the vsetvl instruction encoding. (define_attr "sew" "" - (cond [(eq_attr "mode" "RVVMF64BI,RVVMF32BI,RVVMF16BI,RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ + (cond [(eq_attr "mode" "RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ RVVM8QI,RVVM4QI,RVVM2QI,RVVM1QI,RVVMF2QI,RVVMF4QI,RVVMF8QI,\ RVVM1x8QI,RVVMF2x8QI,RVVMF4x8QI,RVVMF8x8QI,\ RVVM1x7QI,RVVMF2x7QI,RVVMF4x7QI,RVVMF8x7QI,\ @@ -99,6 +99,18 @@ V1QI,V2QI,V4QI,V8QI,V16QI,V32QI,V64QI,V128QI,V256QI,V512QI,V1024QI,V2048QI,V4096QI,\ V1BI,V2BI,V4BI,V8BI,V16BI,V32BI,V64BI,V128BI,V256BI,V512BI,V1024BI,V2048BI,V4096BI") (const_int 8) + (eq_attr "mode" "RVVMF16BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 16) + (const_int 8)) + (eq_attr "mode" "RVVMF32BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 32) + (const_int 8)) + (eq_attr "mode" "RVVMF64BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 64) + (const_int 8)) (eq_attr "mode" "RVVM8HI,RVVM4HI,RVVM2HI,RVVM1HI,RVVMF2HI,RVVMF4HI,\ RVVM1x8HI,RVVMF2x8HI,RVVMF4x8HI,\ RVVM1x7HI,RVVMF2x7HI,RVVMF4x7HI,\ @@ -159,9 +171,9 @@ (eq_attr "mode" "RVVM4QI,RVVMF2BI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2QI,RVVMF4BI") (symbol_ref "riscv_vector::LMUL_2") (eq_attr "mode" "RVVM1QI,RVVMF8BI") (symbol_ref "riscv_vector::LMUL_1") - (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "riscv_vector::LMUL_F2") - (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "riscv_vector::LMUL_F4") - (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "riscv_vector::LMUL_F8") + (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F2") + (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F4") + (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F8") (eq_attr "mode" "RVVM8HI") (symbol_ref "riscv_vector::LMUL_8") (eq_attr "mode" "RVVM4HI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2HI") (symbol_ref "riscv_vector::LMUL_2") @@ -436,6 +448,10 @@ vaeskf1,vaeskf2,vaesz,vsha2ms,vsha2ch,vsha2cl,vsm4k,vsm4r,\ vsm3me,vsm3c") (const_int INVALID_ATTRIBUTE) + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE) (eq_attr "mode" "RVVM8QI,RVVM1BI") (const_int 1) (eq_attr "mode" "RVVM4QI,RVVMF2BI") (const_int 2) (eq_attr "mode" "RVVM2QI,RVVMF4BI") (const_int 4) @@ -905,6 +921,8 @@ (symbol_ref "riscv_vector::FRM_DYN")] (symbol_ref "riscv_vector::FRM_NONE"))) +(include "thead-vector.md") + ;; ----------------------------------------------------------------- ;; ---- Miscellaneous Operations ;; ----------------------------------------------------------------- @@ -1078,6 +1096,12 @@ - We can not leave it to TARGET_SECONDARY_RELOAD since it happens before spilling. The clobber scratch is used by spilling fractional registers in IRA/LRA so it's too early. */ + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; @@ -1114,7 +1138,7 @@ (define_insn "*mov<mode>_whole" [(set (match_operand:V_WHOLE 0 "reg_or_mem_operand" "=vr, m,vr") (match_operand:V_WHOLE 1 "reg_or_mem_operand" " m,vr,vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "@ vl%m1re<sew>.v\t%0,%1 vs%m1r.v\t%1,%0 @@ -1135,6 +1159,13 @@ (match_operand:VB 1 "general_operand"))] "TARGET_VECTOR" { + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } + if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; }) @@ -1142,7 +1173,7 @@ (define_insn "*mov<mode>" [(set (match_operand:VB 0 "register_operand" "=vr") (match_operand:VB 1 "register_operand" " vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "vmv1r.v\t%0,%1" [(set_attr "type" "vmov") (set_attr "mode" "<MODE>")]) @@ -3692,7 +3723,7 @@ (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3713,7 +3744,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3734,7 +3765,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c index 2e0e12aa045..2eef9e1e1a8 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c @@ -1,4 +1,4 @@ -/* { dg-do compile } */ +/* { dg-do compile { target { ! riscv_xtheadvector } } } */ /* { dg-skip-if "test rvv intrinsic" { *-*-* } { "*" } { "-march=rv*v*" } } */ void foo0 () {__rvv_bool64_t t;} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c index 3d81b179235..ef329e30785 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c @@ -1,4 +1,4 @@ /* { dg-do compile } */ /* { dg-options "-O3 -march=rv32gc -mabi=ilp32d" } */ -#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' extension enabled} } */ +#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' or 'XTHEADVECTOR' extension enabled} } */ diff --git a/gcc/testsuite/lib/target-supports.exp b/gcc/testsuite/lib/target-supports.exp index b27c30b8c51..bf0ad5342d6 100644 --- a/gcc/testsuite/lib/target-supports.exp +++ b/gcc/testsuite/lib/target-supports.exp @@ -1956,6 +1956,18 @@ proc check_effective_target_riscv_zbb { } { }] } +# Return 1 if the target arch supports the XTheadVector extension, 0 otherwise. +# Cache the result. + +proc check_effective_target_riscv_xtheadvector { } { + return [check_no_compiler_messages riscv_ext_xtheadvector assembly { + #ifndef __riscv_xtheadvector + #error "Not __riscv_xtheadvector" + #endif + }] +} + + # Return 1 if we can execute code when using dg-add-options riscv_v proc check_effective_target_riscv_v_ok { } { -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* 回复：Re：[PATCH v5] RISC-V: Handle differences between XTheadvector and Vector 2024-01-10 7:17 ` Re：[PATCH " juzhe.zhong @ 2024-01-10 7:26 ` joshua 2024-01-10 7:31 ` 回复：[PATCH " juzhe.zhong 2024-01-10 7:28 ` Re：Re：[PATCH " joshua 1 sibling, 1 reply; 133+ messages in thread From: joshua @ 2024-01-10 7:26 UTC (permalink / raw) To: juzhe.zhong, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu For example +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + /* Report an error if there is no xtheadvector. */ + if (!TARGET_XTHEADVECTOR) + return; + + build_all (b, group); + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Report an error if there is no xtheadvector. */ + if (!TARGET_XTHEADVECTOR) + return nullptr; + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_base_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; I cannot find totally the sam shape that I can reuse. Maybe loadstore_def? But we do not need to do vop --> vop<sew> for our new intrinsics. If we reuse this shape, we need to add some logic here. Also, the shape "th_extract" for "ext" is a new shape that existing shapes haven't implemented. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月10日(星期三) 15:17 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Handle differences between XTheadvector and Vector Why do you need to invade existing shapes ? juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-10 15:16 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Handle differences between XTheadvector and Vector These xttheadvector speical intrinsics are different from rvv1.0 in determining function name from base name. We cannot directly reuse the existing shapes. In order not to invade existing shapes, we add new shapes for new functions. Also, we create new thead-vector-builtins.cc for xtheadvector function_base implementation. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月10日(星期三) 15:01 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Handle differences between XTheadvector and Vector Why do you add theadvector shapes ? I think you can reuse the current existing shapes. +thead-vector-builtins.o: \+ $(srcdir)/config/riscv/thead-vector-builtins.cc \+ $(CONFIG_H) $(SYSTEM_H) coretypes.h $(TM_H) $(TREE_H) $(RTL_H) \+ $(TM_P_H) memmodel.h insn-codes.h $(OPTABS_H) $(RECOG_H) \+ $(EXPR_H) $(BASIC_BLOCK_H) $(FUNCTION_H) fold-const.h $(GIMPLE_H) \+ gimple-iterator.h gimplify.h explow.h $(EMIT_RTL_H) tree-vector-builder.h \+ rtx-vector-builder.h \+ $(srcdir)/config/riscv/riscv-vector-builtins-shapes.h \+ $(srcdir)/config/riscv/riscv-vector-builtins-bases.h \+ $(srcdir)/config/riscv/thead-vector-builtins.h \+ $(RISCV_BUILTINS_H)+ $(COMPILER) -c $(ALL_COMPILERFLAGS) $(ALL_CPPFLAGS) $(INCLUDES) \+ $(srcdir)/config/riscv/thead-vector-builtins.cc+ Why do you rebuild another new object ? + Copyright (C) 2022-2023 Free Software Foundation, Inc. Incorrect copyright juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-10 10:57 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Handle differences between XTheadvector and Vector Hi Juzhe, Thank you for so many useful comments for this patch! There are some more patches to support xtheadvector special instrinsics as well as handle register overlap issue and rewrite assembly output. https://gcc.gnu.org/pipermail/gcc-patches/2024-January/641774.html https://gcc.gnu.org/pipermail/gcc-patches/2024-January/641732.html https://gcc.gnu.org/pipermail/gcc-patches/2024-January/641733.html Also, there is a precedent patch to refactor riscv-vector-builtins-bases.cc https://gcc.gnu.org/pipermail/gcc-patches/2023-December/641530.html Jeff has reviewed it, but didn't have LGTM yet. Joshua ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月10日(星期三) 10:34 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Handle differences between XTheadvector and Vector Thanks for your patience. LGTM from myside. I think it's pretty clean now. I can image in the future when some day the theadvector is no longer used, we can remove it very easily. And also, the theadvector won't affect our RVV1.0 maintain since it's isolated cleanly. But I'd like to wait for a few more days some body want to chime in. And you should do more things before commit it: 1. Remember you should run the full coverage RVV1.0 API test, the test-generator is downloaded from official intrinsic doc: https://github.com/riscv-non-isa/rvv-intrinsic-doc 2. Also the regression of RV32 an RV64 of GCC testsuite. Do you have more patches of theadvector that I didn't review ? plz point them to me again. Thanks. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-10 10:22 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Handle differences between XTheadvector and Vector This patch is to handle the differences in instruction generation between Vector and XTheadVector. In this version, we only support partial xtheadvector instructions that leverage directly from current RVV1.0 with simple adding "th." prefix. For different name xtheadvector instructions but share same patterns as RVV1.0 instructions, we will use ASM targethook to rewrite the whole string of the instructions in the following patches. For some vector patterns that cannot be avoided, we use "!TARGET_XTHEADVECTOR" to disable them in vector.md in order not to generate instructions that xtheadvector does not support, like vmv1r and vsext.vf2. gcc/ChangeLog: * config.gcc: Add files for XTheadVector intrinsics. * config/riscv/autovec.md: Guard XTheadVector. * config/riscv/riscv-c.cc: Add pragma for XTheadVector. * config/riscv/riscv-string.cc (expand_block_move): Guard XTheadVector. * config/riscv/riscv-string.cc (vls_mode_valid_p): Avoid autovec. * config/riscv/riscv-vector-builtins-shapes.cc (check_type): (build_one): New function. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): (DEF_THEAD_RVV_FUNCTION): Add new marcos. (check_required_extensions): (handle_pragma_vector): * config/riscv/riscv-vector-builtins.h (RVV_REQUIRE_VECTOR): (RVV_REQUIRE_XTHEADVECTOR): Add RVV_REQUIRE_VECTOR and RVV_REQUIRE_XTHEADVECTOR. (struct function_group_info): * config/riscv/riscv-vector-switch.def (ENTRY): Disable fractional mode for the XTheadVector extension. (TUPLE_ENTRY): Likewise. * config/riscv/riscv-vsetvl.cc: Add functions for xtheadvector. * config/riscv/riscv.cc (riscv_v_ext_vls_mode_p): Guard XTheadVector. (riscv_v_adjust_bytesize): Likewise. (riscv_preferred_simd_mode): Likewsie. (riscv_autovectorize_vector_modes): Likewise. (riscv_vector_mode_supported_any_target_p): Likewise. (TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P): Likewise. * config/riscv/vector.md: Include thead-vector.md. * config/riscv/riscv_th_vector.h: New file. * config/riscv/thead-vector.md: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/base/pragma-1.c: Add XTheadVector. * gcc.target/riscv/rvv/base/abi-1.c: Exclude XTheadVector. * lib/target-supports.exp: Add target for XTheadVector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config.gcc | 2 +- gcc/config/riscv/autovec.md | 2 +- gcc/config/riscv/predicates.md | 4 +- gcc/config/riscv/riscv-c.cc | 3 +- gcc/config/riscv/riscv-string.cc | 3 +- gcc/config/riscv/riscv-v.cc | 2 +- .../riscv/riscv-vector-builtins-bases.cc | 48 ++++-- .../riscv/riscv-vector-builtins-shapes.cc | 23 +++ gcc/config/riscv/riscv-vector-switch.def | 150 +++++++++--------- gcc/config/riscv/riscv.cc | 20 ++- gcc/config/riscv/riscv_th_vector.h | 49 ++++++ gcc/config/riscv/thead-vector.md | 102 ++++++++++++ gcc/config/riscv/thead.cc | 23 ++- gcc/config/riscv/vector.md | 49 ++++-- .../gcc.target/riscv/rvv/base/abi-1.c | 2 +- .../gcc.target/riscv/rvv/base/pragma-1.c | 2 +- gcc/testsuite/lib/target-supports.exp | 12 ++ 17 files changed, 383 insertions(+), 113 deletions(-) create mode 100644 gcc/config/riscv/riscv_th_vector.h create mode 100644 gcc/config/riscv/thead-vector.md diff --git a/gcc/config.gcc b/gcc/config.gcc index 7e583390024..047e4c02cf4 100644 --- a/gcc/config.gcc +++ b/gcc/config.gcc @@ -549,7 +549,7 @@ riscv*) extra_objs="${extra_objs} riscv-vector-builtins.o riscv-vector-builtins-shapes.o riscv-vector-builtins-bases.o" extra_objs="${extra_objs} thead.o riscv-target-attr.o" d_target_objs="riscv-d.o" - extra_headers="riscv_vector.h" + extra_headers="riscv_vector.h riscv_th_vector.h" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.cc" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.h" ;; diff --git a/gcc/config/riscv/autovec.md b/gcc/config/riscv/autovec.md index 775eaa825b0..0477781cabe 100644 --- a/gcc/config/riscv/autovec.md +++ b/gcc/config/riscv/autovec.md @@ -2579,7 +2579,7 @@ [(match_operand 0 "register_operand") (match_operand 1 "memory_operand") (match_operand:ANYI 2 "const_int_operand")] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" { riscv_vector::expand_rawmemchr(<MODE>mode, operands[0], operands[1], operands[2]); diff --git a/gcc/config/riscv/predicates.md b/gcc/config/riscv/predicates.md index b1a79cae50a..0337da88284 100644 --- a/gcc/config/riscv/predicates.md +++ b/gcc/config/riscv/predicates.md @@ -428,7 +428,9 @@ ;; Predicates for the V extension. (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (ior (match_test "TARGET_XTHEADVECTOR && rtx_equal_p (op, const0_rtx)") + (match_test "!TARGET_XTHEADVECTOR")) + (match_operand 0 "const_csr_operand")))) (define_special_predicate "autovec_length_operand" (ior (match_operand 0 "pmode_register_operand") diff --git a/gcc/config/riscv/riscv-c.cc b/gcc/config/riscv/riscv-c.cc index 422ddc2c308..2e306057347 100644 --- a/gcc/config/riscv/riscv-c.cc +++ b/gcc/config/riscv/riscv-c.cc @@ -195,7 +195,8 @@ riscv_pragma_intrinsic (cpp_reader *) const char *name = TREE_STRING_POINTER (x); - if (strcmp (name, "vector") == 0) + if (strcmp (name, "vector") == 0 + || strcmp (name, "xtheadvector") == 0) { if (!TARGET_VECTOR) { diff --git a/gcc/config/riscv/riscv-string.cc b/gcc/config/riscv/riscv-string.cc index f227b8667ce..b09b51d7526 100644 --- a/gcc/config/riscv/riscv-string.cc +++ b/gcc/config/riscv/riscv-string.cc @@ -773,7 +773,8 @@ riscv_expand_block_move_scalar (rtx dest, rtx src, rtx length) bool riscv_expand_block_move (rtx dest, rtx src, rtx length) { - if (TARGET_VECTOR && stringop_strategy & STRATEGY_VECTOR) + if ((TARGET_VECTOR && !TARGET_XTHEADVECTOR) + && stringop_strategy & STRATEGY_VECTOR) { bool ok = riscv_vector::expand_block_move (dest, src, length); if (ok) diff --git a/gcc/config/riscv/riscv-v.cc b/gcc/config/riscv/riscv-v.cc index 2491522191a..7e25deb966d 100644 --- a/gcc/config/riscv/riscv-v.cc +++ b/gcc/config/riscv/riscv-v.cc @@ -4392,7 +4392,7 @@ cmp_lmul_gt_one (machine_mode mode) bool vls_mode_valid_p (machine_mode vls_mode) { - if (!TARGET_VECTOR) + if (!TARGET_VECTOR || TARGET_XTHEADVECTOR) return false; if (riscv_autovec_preference == RVV_SCALABLE) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 91af2c00674..46f1a1da33e 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -115,23 +115,37 @@ public: tree type = builtin_types[e.type.index].vector; machine_mode mode = TYPE_MODE (type); - /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. - - - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 - - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 - - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 - - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 - - e8,m2/e16,m4/e32,m8 --> e8m2 - - e8,m4/e16,m8 --> e8m4 - - e8,m8 --> e8m8 - */ - /* SEW. */ - e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); - - /* LMUL. */ - machine_mode e8_mode - = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); - e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + + if (TARGET_XTHEADVECTOR) + { + machine_mode inner_mode = GET_MODE_INNER (mode); + /* SEW. */ + e.add_input_operand (Pmode, + gen_int_mode (GET_MODE_BITSIZE (inner_mode), Pmode)); + /* LMUL. */ + e.add_input_operand (Pmode, + gen_int_mode (get_vlmul (mode), Pmode)); + } + else + { + /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. + + - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 + - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 + - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 + - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 + - e8,m2/e16,m4/e32,m8 --> e8m2 + - e8,m4/e16,m8 --> e8m4 + - e8,m8 --> e8m8 + */ + /* SEW. */ + e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); + + /* LMUL. */ + machine_mode e8_mode + = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); + e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + } /* TAIL_ANY. */ e.add_input_operand (Pmode, diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index ee8058dcdb1..1e4f4d53de6 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -33,6 +33,25 @@ namespace riscv_vector { +/* Check whether the RETURN_TYPE and ARGUMENT_TYPES are + valid for the function. */ + +static bool +check_type (tree return_type, vec<tree> &argument_types) +{ + tree arg; + unsigned i; + + if (!return_type) + return false; + + FOR_EACH_VEC_ELT (argument_types, i, arg) + if (!arg) + return false; + + return true; +} + /* Add one function instance for GROUP, using operand suffix at index OI, mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ static void @@ -49,6 +68,10 @@ build_one (function_builder &b, const function_group_info &group, group.ops_infos.types[vec_type_idx].index); b.allocate_argument_types (function_instance, argument_types); b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + b.add_overloaded_function (function_instance, *group.shape); b.add_unique_function (function_instance, (*group.shape), return_type, argument_types); diff --git a/gcc/config/riscv/riscv-vector-switch.def b/gcc/config/riscv/riscv-vector-switch.def index 1ad26c2a3b2..452283b7416 100644 --- a/gcc/config/riscv/riscv-vector-switch.def +++ b/gcc/config/riscv/riscv-vector-switch.def @@ -68,9 +68,9 @@ Encode the ratio of SEW/LMUL into the mask types. #endif /* Disable modes if TARGET_MIN_VLEN == 32. */ -ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) -ENTRY (RVVMF32BI, true, LMUL_F4, 32) -ENTRY (RVVMF16BI, true, LMUL_F2, 16) +ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F8, 64) +ENTRY (RVVMF32BI, true, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F4, 32) +ENTRY (RVVMF16BI, true, TARGET_XTHEADVECTOR ? LMUL_1 : LMUL_F2 , 16) ENTRY (RVVMF8BI, true, LMUL_1, 8) ENTRY (RVVMF4BI, true, LMUL_2, 4) ENTRY (RVVMF2BI, true, LMUL_4, 2) @@ -81,39 +81,39 @@ ENTRY (RVVM8QI, true, LMUL_8, 1) ENTRY (RVVM4QI, true, LMUL_4, 2) ENTRY (RVVM2QI, true, LMUL_2, 4) ENTRY (RVVM1QI, true, LMUL_1, 8) -ENTRY (RVVMF2QI, true, LMUL_F2, 16) -ENTRY (RVVMF4QI, true, LMUL_F4, 32) -ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) +ENTRY (RVVMF2QI, !TARGET_XTHEADVECTOR, LMUL_F2, 16) +ENTRY (RVVMF4QI, !TARGET_XTHEADVECTOR, LMUL_F4, 32) +ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F8, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8HI, true, LMUL_8, 2) ENTRY (RVVM4HI, true, LMUL_4, 4) ENTRY (RVVM2HI, true, LMUL_2, 8) ENTRY (RVVM1HI, true, LMUL_1, 16) -ENTRY (RVVMF2HI, true, LMUL_F2, 32) -ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HI, !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_16. */ ENTRY (RVVM8HF, TARGET_VECTOR_ELEN_FP_16, LMUL_8, 2) ENTRY (RVVM4HF, TARGET_VECTOR_ELEN_FP_16, LMUL_4, 4) ENTRY (RVVM2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_2, 8) ENTRY (RVVM1HF, TARGET_VECTOR_ELEN_FP_16, LMUL_1, 16) -ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_F2, 32) -ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8SI, true, LMUL_8, 4) ENTRY (RVVM4SI, true, LMUL_4, 8) ENTRY (RVVM2SI, true, LMUL_2, 16) ENTRY (RVVM1SI, true, LMUL_1, 32) -ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_32. */ ENTRY (RVVM8SF, TARGET_VECTOR_ELEN_FP_32, LMUL_8, 4) ENTRY (RVVM4SF, TARGET_VECTOR_ELEN_FP_32, LMUL_4, 8) ENTRY (RVVM2SF, TARGET_VECTOR_ELEN_FP_32, LMUL_2, 16) ENTRY (RVVM1SF, TARGET_VECTOR_ELEN_FP_32, LMUL_1, 32) -ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if !TARGET_VECTOR_ELEN_64. */ ENTRY (RVVM8DI, TARGET_VECTOR_ELEN_64, LMUL_8, 8) @@ -140,127 +140,127 @@ ENTRY (RVVM1DF, TARGET_VECTOR_ELEN_FP_64, LMUL_1, 64) #endif TUPLE_ENTRY (RVVM1x8QI, true, RVVM1QI, 8, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x8QI, true, RVVMF2QI, 8, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x8QI, true, RVVMF4QI, 8, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 8, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x8QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 8, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x8QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 8, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 8, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x7QI, true, RVVM1QI, 7, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x7QI, true, RVVMF2QI, 7, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x7QI, true, RVVMF4QI, 7, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 7, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x7QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 7, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x7QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 7, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 7, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x6QI, true, RVVM1QI, 6, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x6QI, true, RVVMF2QI, 6, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x6QI, true, RVVMF4QI, 6, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 6, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x6QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 6, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x6QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 6, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 6, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x5QI, true, RVVM1QI, 5, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x5QI, true, RVVMF2QI, 5, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x5QI, true, RVVMF4QI, 5, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 5, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x5QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 5, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x5QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 5, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 5, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x4QI, true, RVVM2QI, 4, LMUL_2, 4) TUPLE_ENTRY (RVVM1x4QI, true, RVVM1QI, 4, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x4QI, true, RVVMF2QI, 4, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x4QI, true, RVVMF4QI, 4, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 4, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x4QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 4, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x4QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 4, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 4, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x3QI, true, RVVM2QI, 3, LMUL_2, 4) TUPLE_ENTRY (RVVM1x3QI, true, RVVM1QI, 3, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x3QI, true, RVVMF2QI, 3, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x3QI, true, RVVMF4QI, 3, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 3, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x3QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 3, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x3QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 3, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 3, LMUL_F8, 64) TUPLE_ENTRY (RVVM4x2QI, true, RVVM4QI, 2, LMUL_4, 2) TUPLE_ENTRY (RVVM2x2QI, true, RVVM2QI, 2, LMUL_2, 4) TUPLE_ENTRY (RVVM1x2QI, true, RVVM1QI, 2, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x2QI, true, RVVMF2QI, 2, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x2QI, true, RVVMF4QI, 2, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 2, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x2QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 2, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x2QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 2, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 2, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x8HI, true, RVVM1HI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HI, true, RVVMF2HI, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HI, true, RVVM1HI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HI, true, RVVMF2HI, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HI, true, RVVM1HI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HI, true, RVVMF2HI, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HI, true, RVVM1HI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HI, true, RVVMF2HI, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HI, true, RVVM2HI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HI, true, RVVM1HI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HI, true, RVVMF2HI, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HI, true, RVVM2HI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HI, true, RVVM1HI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HI, true, RVVMF2HI, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HI, true, RVVM4HI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HI, true, RVVM2HI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HI, true, RVVM1HI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HI, true, RVVMF2HI, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM4HF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8SI, true, RVVM1SI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SI, true, RVVM1SI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SI, true, RVVM1SI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SI, true, RVVM1SI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SI, true, RVVM2SI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SI, true, RVVM1SI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SI, true, RVVM2SI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SI, true, RVVM1SI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SI, true, RVVM4SI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SI, true, RVVM2SI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SI, true, RVVM1SI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM4SF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 8, LMUL_1, 16) TUPLE_ENTRY (RVVM1x7DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 7, LMUL_1, 16) diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index 8f74cd2e502..28cf1f414e9 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -1406,6 +1406,9 @@ riscv_v_adjust_bytesize (machine_mode mode, int scale) { if (riscv_v_ext_vector_mode_p (mode)) { + if (TARGET_XTHEADVECTOR) + return BYTES_PER_RISCV_VECTOR; + poly_int64 nunits = GET_MODE_NUNITS (mode); poly_int64 mode_size = GET_MODE_SIZE (mode); @@ -9995,7 +9998,7 @@ riscv_use_divmod_expander (void) static machine_mode riscv_preferred_simd_mode (scalar_mode mode) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::preferred_simd_mode (mode); return word_mode; @@ -10346,7 +10349,7 @@ riscv_mode_priority (int, int n) unsigned int riscv_autovectorize_vector_modes (vector_modes *modes, bool all) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::autovectorize_vector_modes (modes, all); return default_autovectorize_vector_modes (modes, all); @@ -10529,6 +10532,16 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) return false; } +/* Implements target hook vector_mode_supported_any_target_p. */ + +static bool +riscv_vector_mode_supported_any_target_p (machine_mode mode) +{ + if (TARGET_XTHEADVECTOR) + return false; + return true; +} + /* Initialize the GCC target structure. */ #undef TARGET_ASM_ALIGNED_HI_OP #define TARGET_ASM_ALIGNED_HI_OP "\t.half\t" @@ -10872,6 +10885,9 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) #undef TARGET_PREFERRED_ELSE_VALUE #define TARGET_PREFERRED_ELSE_VALUE riscv_preferred_else_value +#undef TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P +#define TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P riscv_vector_mode_supported_any_target_p + struct gcc_target targetm = TARGET_INITIALIZER; #include "gt-riscv.h" diff --git a/gcc/config/riscv/riscv_th_vector.h b/gcc/config/riscv/riscv_th_vector.h new file mode 100644 index 00000000000..b6b6738bdda --- /dev/null +++ b/gcc/config/riscv/riscv_th_vector.h @@ -0,0 +1,49 @@ +/* RISC-V 'XTheadVector' Extension intrinsics include file. + Copyright (C) 2024 Free Software Foundation, Inc. + + This file is part of GCC. + + GCC is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published + by the Free Software Foundation; either version 3, or (at your + option) any later version. + + GCC is distributed in the hope that it will be useful, but WITHOUT + ANY WARRANTY; without even the implied warranty of MERCHANTABILITY + or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public + License for more details. + + Under Section 7 of GPL version 3, you are granted additional + permissions described in the GCC Runtime Library Exception, version + 3.1, as published by the Free Software Foundation. + + You should have received a copy of the GNU General Public License and + a copy of the GCC Runtime Library Exception along with this program; + see the files COPYING3 and COPYING.RUNTIME respectively. If not, see + <" target="_blank">" target="_blank">" target="_blank">http://www.gnu.org/licenses/>. */ + +#ifndef __RISCV_TH_VECTOR_H +#define __RISCV_TH_VECTOR_H + +#include <stdint.h> +#include <stddef.h> + +#ifndef __riscv_xtheadvector +#error "XTheadVector intrinsics require the xtheadvector extension." +#else +#ifdef __cplusplus +extern "C" { +#endif + +/* NOTE: This implementation of riscv_th_vector.h is intentionally short. It does + not define the RVV types and intrinsic functions directly in C and C++ + code, but instead uses the following pragma to tell GCC to insert the + necessary type and function definitions itself. The net effect is the + same, and the file is a complete implementation of riscv_th_vector.h. */ +#pragma riscv intrinsic "xtheadvector" + +#ifdef __cplusplus +} +#endif // __cplusplus +#endif // __riscv_xtheadvector +#endif // __RISCV_TH_ECTOR_H diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md new file mode 100644 index 00000000000..696b815252d --- /dev/null +++ b/gcc/config/riscv/thead-vector.md @@ -0,0 +1,102 @@ +(define_c_enum "unspec" [ + UNSPEC_TH_VWLDST +]) + +(define_mode_iterator V_VLS_VT [V VLS VT]) +(define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) + +(define_split + [(set (match_operand:V_VB_VLS_VT 0 "reg_or_mem_operand") + (match_operand:V_VB_VLS_VT 1 "reg_or_mem_operand"))] + "TARGET_XTHEADVECTOR" + [(const_int 0)] + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + }) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:V_VLS_VT 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:V_VLS_VT + [(match_operand:V_VLS_VT 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2")]) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:VB 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:VB + [(match_operand:VB 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2") + (set (attr "sew") (const_int 8)) + (set (attr "vlmul") (symbol_ref "riscv_vector::LMUL_1"))]) + +(define_insn_and_split "*pred_th_mov<mode>" + [(set (match_operand:VB_VLS 0 "nonimmediate_operand" "=vr, m, vr, vr, vr") + (if_then_else:VB_VLS + (unspec:VB_VLS + [(match_operand:VB_VLS 1 "vector_all_trues_mask_operand" "Wc1, Wc1, Wc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) + (match_operand:VB_VLS 3 "vector_move_operand" " m, vr, vr, Wc0, Wc1") + (match_operand:VB_VLS 2 "vector_undef_operand" " vu, vu, vu, vu, vu")))] + "TARGET_XTHEADVECTOR" + "@ + # + # + vmcpy.m\t%0,%3 + vmclr.m\t%0 + vmset.m\t%0" + "&& !reload_completed" + [(const_int 0)] + { + if ((MEM_P (operands[0]) || MEM_P (operands[3])) + || (REG_P (operands[0]) && REG_P (operands[3]) + && INTVAL (operands[5]) == riscv_vector::VLMAX)) + { + emit_move_insn (operands[0], operands[3]); + DONE; + } + FAIL; + } + [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/config/riscv/thead.cc b/gcc/config/riscv/thead.cc index 00b2e14c6c0..cc04101b256 100644 --- a/gcc/config/riscv/thead.cc +++ b/gcc/config/riscv/thead.cc @@ -27,6 +27,7 @@ #include "backend.h" #include "tree.h" #include "rtl.h" +#include "insn-attr.h" #include "explow.h" #include "memmodel.h" #include "emit-rtl.h" @@ -890,8 +891,26 @@ th_asm_output_opcode (FILE *asm_out_file, const char *p) { /* We need to add th. prefix to all the xtheadvector instructions here.*/ - if (current_output_insn != NULL && p[0] == 'v') - fputs ("th.", asm_out_file); + if (current_output_insn != NULL) + { + if (get_attr_type (current_output_insn) == TYPE_VSETVL) + { + if (strstr (p, "zero")) + { + if (strstr (p, "zero,zero")) + return "th.vsetvli\tzero,zero,e%0,%m1"; + else + return "th.vsetvli\tzero,%0,e%1,%m2"; + } + else + { + return "th.vsetvli\t%0,%1,e%2,%m3"; + } + } + + if (p[0] == 'v') + fputs ("th.", asm_out_file); + } return p; } diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index 24b7b4394be..3eb6daafbc2 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -87,7 +87,7 @@ ;; check. However, we need default value of SEW for vsetvl instruction since there ;; is no field for ratio in the vsetvl instruction encoding. (define_attr "sew" "" - (cond [(eq_attr "mode" "RVVMF64BI,RVVMF32BI,RVVMF16BI,RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ + (cond [(eq_attr "mode" "RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ RVVM8QI,RVVM4QI,RVVM2QI,RVVM1QI,RVVMF2QI,RVVMF4QI,RVVMF8QI,\ RVVM1x8QI,RVVMF2x8QI,RVVMF4x8QI,RVVMF8x8QI,\ RVVM1x7QI,RVVMF2x7QI,RVVMF4x7QI,RVVMF8x7QI,\ @@ -99,6 +99,18 @@ V1QI,V2QI,V4QI,V8QI,V16QI,V32QI,V64QI,V128QI,V256QI,V512QI,V1024QI,V2048QI,V4096QI,\ V1BI,V2BI,V4BI,V8BI,V16BI,V32BI,V64BI,V128BI,V256BI,V512BI,V1024BI,V2048BI,V4096BI") (const_int 8) + (eq_attr "mode" "RVVMF16BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 16) + (const_int 8)) + (eq_attr "mode" "RVVMF32BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 32) + (const_int 8)) + (eq_attr "mode" "RVVMF64BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 64) + (const_int 8)) (eq_attr "mode" "RVVM8HI,RVVM4HI,RVVM2HI,RVVM1HI,RVVMF2HI,RVVMF4HI,\ RVVM1x8HI,RVVMF2x8HI,RVVMF4x8HI,\ RVVM1x7HI,RVVMF2x7HI,RVVMF4x7HI,\ @@ -159,9 +171,9 @@ (eq_attr "mode" "RVVM4QI,RVVMF2BI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2QI,RVVMF4BI") (symbol_ref "riscv_vector::LMUL_2") (eq_attr "mode" "RVVM1QI,RVVMF8BI") (symbol_ref "riscv_vector::LMUL_1") - (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "riscv_vector::LMUL_F2") - (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "riscv_vector::LMUL_F4") - (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "riscv_vector::LMUL_F8") + (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F2") + (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F4") + (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F8") (eq_attr "mode" "RVVM8HI") (symbol_ref "riscv_vector::LMUL_8") (eq_attr "mode" "RVVM4HI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2HI") (symbol_ref "riscv_vector::LMUL_2") @@ -436,6 +448,10 @@ vaeskf1,vaeskf2,vaesz,vsha2ms,vsha2ch,vsha2cl,vsm4k,vsm4r,\ vsm3me,vsm3c") (const_int INVALID_ATTRIBUTE) + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE) (eq_attr "mode" "RVVM8QI,RVVM1BI") (const_int 1) (eq_attr "mode" "RVVM4QI,RVVMF2BI") (const_int 2) (eq_attr "mode" "RVVM2QI,RVVMF4BI") (const_int 4) @@ -905,6 +921,8 @@ (symbol_ref "riscv_vector::FRM_DYN")] (symbol_ref "riscv_vector::FRM_NONE"))) +(include "thead-vector.md") + ;; ----------------------------------------------------------------- ;; ---- Miscellaneous Operations ;; ----------------------------------------------------------------- @@ -1078,6 +1096,12 @@ - We can not leave it to TARGET_SECONDARY_RELOAD since it happens before spilling. The clobber scratch is used by spilling fractional registers in IRA/LRA so it's too early. */ + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; @@ -1114,7 +1138,7 @@ (define_insn "*mov<mode>_whole" [(set (match_operand:V_WHOLE 0 "reg_or_mem_operand" "=vr, m,vr") (match_operand:V_WHOLE 1 "reg_or_mem_operand" " m,vr,vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "@ vl%m1re<sew>.v\t%0,%1 vs%m1r.v\t%1,%0 @@ -1135,6 +1159,13 @@ (match_operand:VB 1 "general_operand"))] "TARGET_VECTOR" { + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } + if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; }) @@ -1142,7 +1173,7 @@ (define_insn "*mov<mode>" [(set (match_operand:VB 0 "register_operand" "=vr") (match_operand:VB 1 "register_operand" " vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "vmv1r.v\t%0,%1" [(set_attr "type" "vmov") (set_attr "mode" "<MODE>")]) @@ -3692,7 +3723,7 @@ (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3713,7 +3744,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3734,7 +3765,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c index 2e0e12aa045..2eef9e1e1a8 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c @@ -1,4 +1,4 @@ -/* { dg-do compile } */ +/* { dg-do compile { target { ! riscv_xtheadvector } } } */ /* { dg-skip-if "test rvv intrinsic" { *-*-* } { "*" } { "-march=rv*v*" } } */ void foo0 () {__rvv_bool64_t t;} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c index 3d81b179235..ef329e30785 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c @@ -1,4 +1,4 @@ /* { dg-do compile } */ /* { dg-options "-O3 -march=rv32gc -mabi=ilp32d" } */ -#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' extension enabled} } */ +#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' or 'XTHEADVECTOR' extension enabled} } */ diff --git a/gcc/testsuite/lib/target-supports.exp b/gcc/testsuite/lib/target-supports.exp index b27c30b8c51..bf0ad5342d6 100644 --- a/gcc/testsuite/lib/target-supports.exp +++ b/gcc/testsuite/lib/target-supports.exp @@ -1956,6 +1956,18 @@ proc check_effective_target_riscv_zbb { } { }] } +# Return 1 if the target arch supports the XTheadVector extension, 0 otherwise. +# Cache the result. + +proc check_effective_target_riscv_xtheadvector { } { + return [check_no_compiler_messages riscv_ext_xtheadvector assembly { + #ifndef __riscv_xtheadvector + #error "Not __riscv_xtheadvector" + #endif + }] +} + + # Return 1 if we can execute code when using dg-add-options riscv_v proc check_effective_target_riscv_v_ok { } { -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: 回复：[PATCH v5] RISC-V: Handle differences between XTheadvector and Vector 2024-01-10 7:26 ` 回复：Re：[PATCH " joshua @ 2024-01-10 7:31 ` juzhe.zhong 0 siblings, 0 replies; 133+ messages in thread From: juzhe.zhong @ 2024-01-10 7:31 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 53226 bytes --] OK. Make sens to me. + /* Report an error if there is no xtheadvector. */ + if (!TARGET_XTHEADVECTOR) + return nullptr; Remove all these checks. They are redundant. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-10 15:26 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： 回复：Re：[PATCH v5] RISC-V: Handle differences between XTheadvector and Vector For example +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + /* Report an error if there is no xtheadvector. */ + if (!TARGET_XTHEADVECTOR) + return; + + build_all (b, group); + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Report an error if there is no xtheadvector. */ + if (!TARGET_XTHEADVECTOR) + return nullptr; + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_base_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; I cannot find totally the sam shape that I can reuse. Maybe loadstore_def? But we do not need to do vop --> vop<sew> for our new intrinsics. If we reuse this shape, we need to add some logic here. Also, the shape "th_extract" for "ext" is a new shape that existing shapes haven't implemented. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月10日(星期三) 15:17 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Handle differences between XTheadvector and Vector Why do you need to invade existing shapes ? juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-10 15:16 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Handle differences between XTheadvector and Vector These xttheadvector speical intrinsics are different from rvv1.0 in determining function name from base name. We cannot directly reuse the existing shapes. In order not to invade existing shapes, we add new shapes for new functions. Also, we create new thead-vector-builtins.cc for xtheadvector function_base implementation. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月10日(星期三) 15:01 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Handle differences between XTheadvector and Vector Why do you add theadvector shapes ? I think you can reuse the current existing shapes. +thead-vector-builtins.o: \+ $(srcdir)/config/riscv/thead-vector-builtins.cc \+ $(CONFIG_H) $(SYSTEM_H) coretypes.h $(TM_H) $(TREE_H) $(RTL_H) \+ $(TM_P_H) memmodel.h insn-codes.h $(OPTABS_H) $(RECOG_H) \+ $(EXPR_H) $(BASIC_BLOCK_H) $(FUNCTION_H) fold-const.h $(GIMPLE_H) \+ gimple-iterator.h gimplify.h explow.h $(EMIT_RTL_H) tree-vector-builder.h \+ rtx-vector-builder.h \+ $(srcdir)/config/riscv/riscv-vector-builtins-shapes.h \+ $(srcdir)/config/riscv/riscv-vector-builtins-bases.h \+ $(srcdir)/config/riscv/thead-vector-builtins.h \+ $(RISCV_BUILTINS_H)+ $(COMPILER) -c $(ALL_COMPILERFLAGS) $(ALL_CPPFLAGS) $(INCLUDES) \+ $(srcdir)/config/riscv/thead-vector-builtins.cc+ Why do you rebuild another new object ? + Copyright (C) 2022-2023 Free Software Foundation, Inc. Incorrect copyright juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-10 10:57 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Handle differences between XTheadvector and Vector Hi Juzhe, Thank you for so many useful comments for this patch! There are some more patches to support xtheadvector special instrinsics as well as handle register overlap issue and rewrite assembly output. https://gcc.gnu.org/pipermail/gcc-patches/2024-January/641774.html https://gcc.gnu.org/pipermail/gcc-patches/2024-January/641732.html https://gcc.gnu.org/pipermail/gcc-patches/2024-January/641733.html Also, there is a precedent patch to refactor riscv-vector-builtins-bases.cc https://gcc.gnu.org/pipermail/gcc-patches/2023-December/641530.html Jeff has reviewed it, but didn't have LGTM yet. Joshua ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月10日(星期三) 10:34 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Handle differences between XTheadvector and Vector Thanks for your patience. LGTM from myside. I think it's pretty clean now. I can image in the future when some day the theadvector is no longer used, we can remove it very easily. And also, the theadvector won't affect our RVV1.0 maintain since it's isolated cleanly. But I'd like to wait for a few more days some body want to chime in. And you should do more things before commit it: 1. Remember you should run the full coverage RVV1.0 API test, the test-generator is downloaded from official intrinsic doc: https://github.com/riscv-non-isa/rvv-intrinsic-doc 2. Also the regression of RV32 an RV64 of GCC testsuite. Do you have more patches of theadvector that I didn't review ? plz point them to me again. Thanks. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-10 10:22 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Handle differences between XTheadvector and Vector This patch is to handle the differences in instruction generation between Vector and XTheadVector. In this version, we only support partial xtheadvector instructions that leverage directly from current RVV1.0 with simple adding "th." prefix. For different name xtheadvector instructions but share same patterns as RVV1.0 instructions, we will use ASM targethook to rewrite the whole string of the instructions in the following patches. For some vector patterns that cannot be avoided, we use "!TARGET_XTHEADVECTOR" to disable them in vector.md in order not to generate instructions that xtheadvector does not support, like vmv1r and vsext.vf2. gcc/ChangeLog: * config.gcc: Add files for XTheadVector intrinsics. * config/riscv/autovec.md: Guard XTheadVector. * config/riscv/riscv-c.cc: Add pragma for XTheadVector. * config/riscv/riscv-string.cc (expand_block_move): Guard XTheadVector. * config/riscv/riscv-string.cc (vls_mode_valid_p): Avoid autovec. * config/riscv/riscv-vector-builtins-shapes.cc (check_type): (build_one): New function. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): (DEF_THEAD_RVV_FUNCTION): Add new marcos. (check_required_extensions): (handle_pragma_vector): * config/riscv/riscv-vector-builtins.h (RVV_REQUIRE_VECTOR): (RVV_REQUIRE_XTHEADVECTOR): Add RVV_REQUIRE_VECTOR and RVV_REQUIRE_XTHEADVECTOR. (struct function_group_info): * config/riscv/riscv-vector-switch.def (ENTRY): Disable fractional mode for the XTheadVector extension. (TUPLE_ENTRY): Likewise. * config/riscv/riscv-vsetvl.cc: Add functions for xtheadvector. * config/riscv/riscv.cc (riscv_v_ext_vls_mode_p): Guard XTheadVector. (riscv_v_adjust_bytesize): Likewise. (riscv_preferred_simd_mode): Likewsie. (riscv_autovectorize_vector_modes): Likewise. (riscv_vector_mode_supported_any_target_p): Likewise. (TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P): Likewise. * config/riscv/vector.md: Include thead-vector.md. * config/riscv/riscv_th_vector.h: New file. * config/riscv/thead-vector.md: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/base/pragma-1.c: Add XTheadVector. * gcc.target/riscv/rvv/base/abi-1.c: Exclude XTheadVector. * lib/target-supports.exp: Add target for XTheadVector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config.gcc | 2 +- gcc/config/riscv/autovec.md | 2 +- gcc/config/riscv/predicates.md | 4 +- gcc/config/riscv/riscv-c.cc | 3 +- gcc/config/riscv/riscv-string.cc | 3 +- gcc/config/riscv/riscv-v.cc | 2 +- .../riscv/riscv-vector-builtins-bases.cc | 48 ++++-- .../riscv/riscv-vector-builtins-shapes.cc | 23 +++ gcc/config/riscv/riscv-vector-switch.def | 150 +++++++++--------- gcc/config/riscv/riscv.cc | 20 ++- gcc/config/riscv/riscv_th_vector.h | 49 ++++++ gcc/config/riscv/thead-vector.md | 102 ++++++++++++ gcc/config/riscv/thead.cc | 23 ++- gcc/config/riscv/vector.md | 49 ++++-- .../gcc.target/riscv/rvv/base/abi-1.c | 2 +- .../gcc.target/riscv/rvv/base/pragma-1.c | 2 +- gcc/testsuite/lib/target-supports.exp | 12 ++ 17 files changed, 383 insertions(+), 113 deletions(-) create mode 100644 gcc/config/riscv/riscv_th_vector.h create mode 100644 gcc/config/riscv/thead-vector.md diff --git a/gcc/config.gcc b/gcc/config.gcc index 7e583390024..047e4c02cf4 100644 --- a/gcc/config.gcc +++ b/gcc/config.gcc @@ -549,7 +549,7 @@ riscv*) extra_objs="${extra_objs} riscv-vector-builtins.o riscv-vector-builtins-shapes.o riscv-vector-builtins-bases.o" extra_objs="${extra_objs} thead.o riscv-target-attr.o" d_target_objs="riscv-d.o" - extra_headers="riscv_vector.h" + extra_headers="riscv_vector.h riscv_th_vector.h" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.cc" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.h" ;; diff --git a/gcc/config/riscv/autovec.md b/gcc/config/riscv/autovec.md index 775eaa825b0..0477781cabe 100644 --- a/gcc/config/riscv/autovec.md +++ b/gcc/config/riscv/autovec.md @@ -2579,7 +2579,7 @@ [(match_operand 0 "register_operand") (match_operand 1 "memory_operand") (match_operand:ANYI 2 "const_int_operand")] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" { riscv_vector::expand_rawmemchr(<MODE>mode, operands[0], operands[1], operands[2]); diff --git a/gcc/config/riscv/predicates.md b/gcc/config/riscv/predicates.md index b1a79cae50a..0337da88284 100644 --- a/gcc/config/riscv/predicates.md +++ b/gcc/config/riscv/predicates.md @@ -428,7 +428,9 @@ ;; Predicates for the V extension. (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (ior (match_test "TARGET_XTHEADVECTOR && rtx_equal_p (op, const0_rtx)") + (match_test "!TARGET_XTHEADVECTOR")) + (match_operand 0 "const_csr_operand")))) (define_special_predicate "autovec_length_operand" (ior (match_operand 0 "pmode_register_operand") diff --git a/gcc/config/riscv/riscv-c.cc b/gcc/config/riscv/riscv-c.cc index 422ddc2c308..2e306057347 100644 --- a/gcc/config/riscv/riscv-c.cc +++ b/gcc/config/riscv/riscv-c.cc @@ -195,7 +195,8 @@ riscv_pragma_intrinsic (cpp_reader *) const char *name = TREE_STRING_POINTER (x); - if (strcmp (name, "vector") == 0) + if (strcmp (name, "vector") == 0 + || strcmp (name, "xtheadvector") == 0) { if (!TARGET_VECTOR) { diff --git a/gcc/config/riscv/riscv-string.cc b/gcc/config/riscv/riscv-string.cc index f227b8667ce..b09b51d7526 100644 --- a/gcc/config/riscv/riscv-string.cc +++ b/gcc/config/riscv/riscv-string.cc @@ -773,7 +773,8 @@ riscv_expand_block_move_scalar (rtx dest, rtx src, rtx length) bool riscv_expand_block_move (rtx dest, rtx src, rtx length) { - if (TARGET_VECTOR && stringop_strategy & STRATEGY_VECTOR) + if ((TARGET_VECTOR && !TARGET_XTHEADVECTOR) + && stringop_strategy & STRATEGY_VECTOR) { bool ok = riscv_vector::expand_block_move (dest, src, length); if (ok) diff --git a/gcc/config/riscv/riscv-v.cc b/gcc/config/riscv/riscv-v.cc index 2491522191a..7e25deb966d 100644 --- a/gcc/config/riscv/riscv-v.cc +++ b/gcc/config/riscv/riscv-v.cc @@ -4392,7 +4392,7 @@ cmp_lmul_gt_one (machine_mode mode) bool vls_mode_valid_p (machine_mode vls_mode) { - if (!TARGET_VECTOR) + if (!TARGET_VECTOR || TARGET_XTHEADVECTOR) return false; if (riscv_autovec_preference == RVV_SCALABLE) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 91af2c00674..46f1a1da33e 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -115,23 +115,37 @@ public: tree type = builtin_types[e.type.index].vector; machine_mode mode = TYPE_MODE (type); - /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. - - - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 - - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 - - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 - - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 - - e8,m2/e16,m4/e32,m8 --> e8m2 - - e8,m4/e16,m8 --> e8m4 - - e8,m8 --> e8m8 - */ - /* SEW. */ - e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); - - /* LMUL. */ - machine_mode e8_mode - = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); - e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + + if (TARGET_XTHEADVECTOR) + { + machine_mode inner_mode = GET_MODE_INNER (mode); + /* SEW. */ + e.add_input_operand (Pmode, + gen_int_mode (GET_MODE_BITSIZE (inner_mode), Pmode)); + /* LMUL. */ + e.add_input_operand (Pmode, + gen_int_mode (get_vlmul (mode), Pmode)); + } + else + { + /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. + + - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 + - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 + - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 + - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 + - e8,m2/e16,m4/e32,m8 --> e8m2 + - e8,m4/e16,m8 --> e8m4 + - e8,m8 --> e8m8 + */ + /* SEW. */ + e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); + + /* LMUL. */ + machine_mode e8_mode + = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); + e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + } /* TAIL_ANY. */ e.add_input_operand (Pmode, diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index ee8058dcdb1..1e4f4d53de6 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -33,6 +33,25 @@ namespace riscv_vector { +/* Check whether the RETURN_TYPE and ARGUMENT_TYPES are + valid for the function. */ + +static bool +check_type (tree return_type, vec<tree> &argument_types) +{ + tree arg; + unsigned i; + + if (!return_type) + return false; + + FOR_EACH_VEC_ELT (argument_types, i, arg) + if (!arg) + return false; + + return true; +} + /* Add one function instance for GROUP, using operand suffix at index OI, mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ static void @@ -49,6 +68,10 @@ build_one (function_builder &b, const function_group_info &group, group.ops_infos.types[vec_type_idx].index); b.allocate_argument_types (function_instance, argument_types); b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + b.add_overloaded_function (function_instance, *group.shape); b.add_unique_function (function_instance, (*group.shape), return_type, argument_types); diff --git a/gcc/config/riscv/riscv-vector-switch.def b/gcc/config/riscv/riscv-vector-switch.def index 1ad26c2a3b2..452283b7416 100644 --- a/gcc/config/riscv/riscv-vector-switch.def +++ b/gcc/config/riscv/riscv-vector-switch.def @@ -68,9 +68,9 @@ Encode the ratio of SEW/LMUL into the mask types. #endif /* Disable modes if TARGET_MIN_VLEN == 32. */ -ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) -ENTRY (RVVMF32BI, true, LMUL_F4, 32) -ENTRY (RVVMF16BI, true, LMUL_F2, 16) +ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F8, 64) +ENTRY (RVVMF32BI, true, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F4, 32) +ENTRY (RVVMF16BI, true, TARGET_XTHEADVECTOR ? LMUL_1 : LMUL_F2 , 16) ENTRY (RVVMF8BI, true, LMUL_1, 8) ENTRY (RVVMF4BI, true, LMUL_2, 4) ENTRY (RVVMF2BI, true, LMUL_4, 2) @@ -81,39 +81,39 @@ ENTRY (RVVM8QI, true, LMUL_8, 1) ENTRY (RVVM4QI, true, LMUL_4, 2) ENTRY (RVVM2QI, true, LMUL_2, 4) ENTRY (RVVM1QI, true, LMUL_1, 8) -ENTRY (RVVMF2QI, true, LMUL_F2, 16) -ENTRY (RVVMF4QI, true, LMUL_F4, 32) -ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) +ENTRY (RVVMF2QI, !TARGET_XTHEADVECTOR, LMUL_F2, 16) +ENTRY (RVVMF4QI, !TARGET_XTHEADVECTOR, LMUL_F4, 32) +ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F8, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8HI, true, LMUL_8, 2) ENTRY (RVVM4HI, true, LMUL_4, 4) ENTRY (RVVM2HI, true, LMUL_2, 8) ENTRY (RVVM1HI, true, LMUL_1, 16) -ENTRY (RVVMF2HI, true, LMUL_F2, 32) -ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HI, !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_16. */ ENTRY (RVVM8HF, TARGET_VECTOR_ELEN_FP_16, LMUL_8, 2) ENTRY (RVVM4HF, TARGET_VECTOR_ELEN_FP_16, LMUL_4, 4) ENTRY (RVVM2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_2, 8) ENTRY (RVVM1HF, TARGET_VECTOR_ELEN_FP_16, LMUL_1, 16) -ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_F2, 32) -ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8SI, true, LMUL_8, 4) ENTRY (RVVM4SI, true, LMUL_4, 8) ENTRY (RVVM2SI, true, LMUL_2, 16) ENTRY (RVVM1SI, true, LMUL_1, 32) -ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_32. */ ENTRY (RVVM8SF, TARGET_VECTOR_ELEN_FP_32, LMUL_8, 4) ENTRY (RVVM4SF, TARGET_VECTOR_ELEN_FP_32, LMUL_4, 8) ENTRY (RVVM2SF, TARGET_VECTOR_ELEN_FP_32, LMUL_2, 16) ENTRY (RVVM1SF, TARGET_VECTOR_ELEN_FP_32, LMUL_1, 32) -ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if !TARGET_VECTOR_ELEN_64. */ ENTRY (RVVM8DI, TARGET_VECTOR_ELEN_64, LMUL_8, 8) @@ -140,127 +140,127 @@ ENTRY (RVVM1DF, TARGET_VECTOR_ELEN_FP_64, LMUL_1, 64) #endif TUPLE_ENTRY (RVVM1x8QI, true, RVVM1QI, 8, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x8QI, true, RVVMF2QI, 8, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x8QI, true, RVVMF4QI, 8, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 8, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x8QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 8, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x8QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 8, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 8, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x7QI, true, RVVM1QI, 7, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x7QI, true, RVVMF2QI, 7, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x7QI, true, RVVMF4QI, 7, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 7, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x7QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 7, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x7QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 7, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 7, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x6QI, true, RVVM1QI, 6, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x6QI, true, RVVMF2QI, 6, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x6QI, true, RVVMF4QI, 6, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 6, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x6QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 6, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x6QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 6, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 6, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x5QI, true, RVVM1QI, 5, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x5QI, true, RVVMF2QI, 5, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x5QI, true, RVVMF4QI, 5, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 5, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x5QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 5, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x5QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 5, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 5, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x4QI, true, RVVM2QI, 4, LMUL_2, 4) TUPLE_ENTRY (RVVM1x4QI, true, RVVM1QI, 4, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x4QI, true, RVVMF2QI, 4, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x4QI, true, RVVMF4QI, 4, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 4, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x4QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 4, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x4QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 4, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 4, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x3QI, true, RVVM2QI, 3, LMUL_2, 4) TUPLE_ENTRY (RVVM1x3QI, true, RVVM1QI, 3, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x3QI, true, RVVMF2QI, 3, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x3QI, true, RVVMF4QI, 3, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 3, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x3QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 3, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x3QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 3, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 3, LMUL_F8, 64) TUPLE_ENTRY (RVVM4x2QI, true, RVVM4QI, 2, LMUL_4, 2) TUPLE_ENTRY (RVVM2x2QI, true, RVVM2QI, 2, LMUL_2, 4) TUPLE_ENTRY (RVVM1x2QI, true, RVVM1QI, 2, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x2QI, true, RVVMF2QI, 2, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x2QI, true, RVVMF4QI, 2, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 2, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x2QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 2, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x2QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 2, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 2, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x8HI, true, RVVM1HI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HI, true, RVVMF2HI, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HI, true, RVVM1HI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HI, true, RVVMF2HI, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HI, true, RVVM1HI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HI, true, RVVMF2HI, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HI, true, RVVM1HI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HI, true, RVVMF2HI, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HI, true, RVVM2HI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HI, true, RVVM1HI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HI, true, RVVMF2HI, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HI, true, RVVM2HI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HI, true, RVVM1HI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HI, true, RVVMF2HI, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HI, true, RVVM4HI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HI, true, RVVM2HI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HI, true, RVVM1HI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HI, true, RVVMF2HI, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM4HF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8SI, true, RVVM1SI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SI, true, RVVM1SI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SI, true, RVVM1SI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SI, true, RVVM1SI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SI, true, RVVM2SI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SI, true, RVVM1SI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SI, true, RVVM2SI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SI, true, RVVM1SI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SI, true, RVVM4SI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SI, true, RVVM2SI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SI, true, RVVM1SI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM4SF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 8, LMUL_1, 16) TUPLE_ENTRY (RVVM1x7DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 7, LMUL_1, 16) diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index 8f74cd2e502..28cf1f414e9 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -1406,6 +1406,9 @@ riscv_v_adjust_bytesize (machine_mode mode, int scale) { if (riscv_v_ext_vector_mode_p (mode)) { + if (TARGET_XTHEADVECTOR) + return BYTES_PER_RISCV_VECTOR; + poly_int64 nunits = GET_MODE_NUNITS (mode); poly_int64 mode_size = GET_MODE_SIZE (mode); @@ -9995,7 +9998,7 @@ riscv_use_divmod_expander (void) static machine_mode riscv_preferred_simd_mode (scalar_mode mode) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::preferred_simd_mode (mode); return word_mode; @@ -10346,7 +10349,7 @@ riscv_mode_priority (int, int n) unsigned int riscv_autovectorize_vector_modes (vector_modes *modes, bool all) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::autovectorize_vector_modes (modes, all); return default_autovectorize_vector_modes (modes, all); @@ -10529,6 +10532,16 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) return false; } +/* Implements target hook vector_mode_supported_any_target_p. */ + +static bool +riscv_vector_mode_supported_any_target_p (machine_mode mode) +{ + if (TARGET_XTHEADVECTOR) + return false; + return true; +} + /* Initialize the GCC target structure. */ #undef TARGET_ASM_ALIGNED_HI_OP #define TARGET_ASM_ALIGNED_HI_OP "\t.half\t" @@ -10872,6 +10885,9 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) #undef TARGET_PREFERRED_ELSE_VALUE #define TARGET_PREFERRED_ELSE_VALUE riscv_preferred_else_value +#undef TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P +#define TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P riscv_vector_mode_supported_any_target_p + struct gcc_target targetm = TARGET_INITIALIZER; #include "gt-riscv.h" diff --git a/gcc/config/riscv/riscv_th_vector.h b/gcc/config/riscv/riscv_th_vector.h new file mode 100644 index 00000000000..b6b6738bdda --- /dev/null +++ b/gcc/config/riscv/riscv_th_vector.h @@ -0,0 +1,49 @@ +/* RISC-V 'XTheadVector' Extension intrinsics include file. + Copyright (C) 2024 Free Software Foundation, Inc. + + This file is part of GCC. + + GCC is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published + by the Free Software Foundation; either version 3, or (at your + option) any later version. + + GCC is distributed in the hope that it will be useful, but WITHOUT + ANY WARRANTY; without even the implied warranty of MERCHANTABILITY + or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public + License for more details. + + Under Section 7 of GPL version 3, you are granted additional + permissions described in the GCC Runtime Library Exception, version + 3.1, as published by the Free Software Foundation. + + You should have received a copy of the GNU General Public License and + a copy of the GCC Runtime Library Exception along with this program; + see the files COPYING3 and COPYING.RUNTIME respectively. If not, see + <" target="_blank">" target="_blank">" target="_blank">http://www.gnu.org/licenses/>. */ + +#ifndef __RISCV_TH_VECTOR_H +#define __RISCV_TH_VECTOR_H + +#include <stdint.h> +#include <stddef.h> + +#ifndef __riscv_xtheadvector +#error "XTheadVector intrinsics require the xtheadvector extension." +#else +#ifdef __cplusplus +extern "C" { +#endif + +/* NOTE: This implementation of riscv_th_vector.h is intentionally short. It does + not define the RVV types and intrinsic functions directly in C and C++ + code, but instead uses the following pragma to tell GCC to insert the + necessary type and function definitions itself. The net effect is the + same, and the file is a complete implementation of riscv_th_vector.h. */ +#pragma riscv intrinsic "xtheadvector" + +#ifdef __cplusplus +} +#endif // __cplusplus +#endif // __riscv_xtheadvector +#endif // __RISCV_TH_ECTOR_H diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md new file mode 100644 index 00000000000..696b815252d --- /dev/null +++ b/gcc/config/riscv/thead-vector.md @@ -0,0 +1,102 @@ +(define_c_enum "unspec" [ + UNSPEC_TH_VWLDST +]) + +(define_mode_iterator V_VLS_VT [V VLS VT]) +(define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) + +(define_split + [(set (match_operand:V_VB_VLS_VT 0 "reg_or_mem_operand") + (match_operand:V_VB_VLS_VT 1 "reg_or_mem_operand"))] + "TARGET_XTHEADVECTOR" + [(const_int 0)] + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + }) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:V_VLS_VT 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:V_VLS_VT + [(match_operand:V_VLS_VT 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2")]) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:VB 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:VB + [(match_operand:VB 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2") + (set (attr "sew") (const_int 8)) + (set (attr "vlmul") (symbol_ref "riscv_vector::LMUL_1"))]) + +(define_insn_and_split "*pred_th_mov<mode>" + [(set (match_operand:VB_VLS 0 "nonimmediate_operand" "=vr, m, vr, vr, vr") + (if_then_else:VB_VLS + (unspec:VB_VLS + [(match_operand:VB_VLS 1 "vector_all_trues_mask_operand" "Wc1, Wc1, Wc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) + (match_operand:VB_VLS 3 "vector_move_operand" " m, vr, vr, Wc0, Wc1") + (match_operand:VB_VLS 2 "vector_undef_operand" " vu, vu, vu, vu, vu")))] + "TARGET_XTHEADVECTOR" + "@ + # + # + vmcpy.m\t%0,%3 + vmclr.m\t%0 + vmset.m\t%0" + "&& !reload_completed" + [(const_int 0)] + { + if ((MEM_P (operands[0]) || MEM_P (operands[3])) + || (REG_P (operands[0]) && REG_P (operands[3]) + && INTVAL (operands[5]) == riscv_vector::VLMAX)) + { + emit_move_insn (operands[0], operands[3]); + DONE; + } + FAIL; + } + [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/config/riscv/thead.cc b/gcc/config/riscv/thead.cc index 00b2e14c6c0..cc04101b256 100644 --- a/gcc/config/riscv/thead.cc +++ b/gcc/config/riscv/thead.cc @@ -27,6 +27,7 @@ #include "backend.h" #include "tree.h" #include "rtl.h" +#include "insn-attr.h" #include "explow.h" #include "memmodel.h" #include "emit-rtl.h" @@ -890,8 +891,26 @@ th_asm_output_opcode (FILE *asm_out_file, const char *p) { /* We need to add th. prefix to all the xtheadvector instructions here.*/ - if (current_output_insn != NULL && p[0] == 'v') - fputs ("th.", asm_out_file); + if (current_output_insn != NULL) + { + if (get_attr_type (current_output_insn) == TYPE_VSETVL) + { + if (strstr (p, "zero")) + { + if (strstr (p, "zero,zero")) + return "th.vsetvli\tzero,zero,e%0,%m1"; + else + return "th.vsetvli\tzero,%0,e%1,%m2"; + } + else + { + return "th.vsetvli\t%0,%1,e%2,%m3"; + } + } + + if (p[0] == 'v') + fputs ("th.", asm_out_file); + } return p; } diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index 24b7b4394be..3eb6daafbc2 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -87,7 +87,7 @@ ;; check. However, we need default value of SEW for vsetvl instruction since there ;; is no field for ratio in the vsetvl instruction encoding. (define_attr "sew" "" - (cond [(eq_attr "mode" "RVVMF64BI,RVVMF32BI,RVVMF16BI,RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ + (cond [(eq_attr "mode" "RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ RVVM8QI,RVVM4QI,RVVM2QI,RVVM1QI,RVVMF2QI,RVVMF4QI,RVVMF8QI,\ RVVM1x8QI,RVVMF2x8QI,RVVMF4x8QI,RVVMF8x8QI,\ RVVM1x7QI,RVVMF2x7QI,RVVMF4x7QI,RVVMF8x7QI,\ @@ -99,6 +99,18 @@ V1QI,V2QI,V4QI,V8QI,V16QI,V32QI,V64QI,V128QI,V256QI,V512QI,V1024QI,V2048QI,V4096QI,\ V1BI,V2BI,V4BI,V8BI,V16BI,V32BI,V64BI,V128BI,V256BI,V512BI,V1024BI,V2048BI,V4096BI") (const_int 8) + (eq_attr "mode" "RVVMF16BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 16) + (const_int 8)) + (eq_attr "mode" "RVVMF32BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 32) + (const_int 8)) + (eq_attr "mode" "RVVMF64BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 64) + (const_int 8)) (eq_attr "mode" "RVVM8HI,RVVM4HI,RVVM2HI,RVVM1HI,RVVMF2HI,RVVMF4HI,\ RVVM1x8HI,RVVMF2x8HI,RVVMF4x8HI,\ RVVM1x7HI,RVVMF2x7HI,RVVMF4x7HI,\ @@ -159,9 +171,9 @@ (eq_attr "mode" "RVVM4QI,RVVMF2BI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2QI,RVVMF4BI") (symbol_ref "riscv_vector::LMUL_2") (eq_attr "mode" "RVVM1QI,RVVMF8BI") (symbol_ref "riscv_vector::LMUL_1") - (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "riscv_vector::LMUL_F2") - (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "riscv_vector::LMUL_F4") - (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "riscv_vector::LMUL_F8") + (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F2") + (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F4") + (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F8") (eq_attr "mode" "RVVM8HI") (symbol_ref "riscv_vector::LMUL_8") (eq_attr "mode" "RVVM4HI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2HI") (symbol_ref "riscv_vector::LMUL_2") @@ -436,6 +448,10 @@ vaeskf1,vaeskf2,vaesz,vsha2ms,vsha2ch,vsha2cl,vsm4k,vsm4r,\ vsm3me,vsm3c") (const_int INVALID_ATTRIBUTE) + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE) (eq_attr "mode" "RVVM8QI,RVVM1BI") (const_int 1) (eq_attr "mode" "RVVM4QI,RVVMF2BI") (const_int 2) (eq_attr "mode" "RVVM2QI,RVVMF4BI") (const_int 4) @@ -905,6 +921,8 @@ (symbol_ref "riscv_vector::FRM_DYN")] (symbol_ref "riscv_vector::FRM_NONE"))) +(include "thead-vector.md") + ;; ----------------------------------------------------------------- ;; ---- Miscellaneous Operations ;; ----------------------------------------------------------------- @@ -1078,6 +1096,12 @@ - We can not leave it to TARGET_SECONDARY_RELOAD since it happens before spilling. The clobber scratch is used by spilling fractional registers in IRA/LRA so it's too early. */ + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; @@ -1114,7 +1138,7 @@ (define_insn "*mov<mode>_whole" [(set (match_operand:V_WHOLE 0 "reg_or_mem_operand" "=vr, m,vr") (match_operand:V_WHOLE 1 "reg_or_mem_operand" " m,vr,vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "@ vl%m1re<sew>.v\t%0,%1 vs%m1r.v\t%1,%0 @@ -1135,6 +1159,13 @@ (match_operand:VB 1 "general_operand"))] "TARGET_VECTOR" { + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } + if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; }) @@ -1142,7 +1173,7 @@ (define_insn "*mov<mode>" [(set (match_operand:VB 0 "register_operand" "=vr") (match_operand:VB 1 "register_operand" " vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "vmv1r.v\t%0,%1" [(set_attr "type" "vmov") (set_attr "mode" "<MODE>")]) @@ -3692,7 +3723,7 @@ (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3713,7 +3744,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3734,7 +3765,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c index 2e0e12aa045..2eef9e1e1a8 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c @@ -1,4 +1,4 @@ -/* { dg-do compile } */ +/* { dg-do compile { target { ! riscv_xtheadvector } } } */ /* { dg-skip-if "test rvv intrinsic" { *-*-* } { "*" } { "-march=rv*v*" } } */ void foo0 () {__rvv_bool64_t t;} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c index 3d81b179235..ef329e30785 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c @@ -1,4 +1,4 @@ /* { dg-do compile } */ /* { dg-options "-O3 -march=rv32gc -mabi=ilp32d" } */ -#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' extension enabled} } */ +#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' or 'XTHEADVECTOR' extension enabled} } */ diff --git a/gcc/testsuite/lib/target-supports.exp b/gcc/testsuite/lib/target-supports.exp index b27c30b8c51..bf0ad5342d6 100644 --- a/gcc/testsuite/lib/target-supports.exp +++ b/gcc/testsuite/lib/target-supports.exp @@ -1956,6 +1956,18 @@ proc check_effective_target_riscv_zbb { } { }] } +# Return 1 if the target arch supports the XTheadVector extension, 0 otherwise. +# Cache the result. + +proc check_effective_target_riscv_xtheadvector { } { + return [check_no_compiler_messages riscv_ext_xtheadvector assembly { + #ifndef __riscv_xtheadvector + #error "Not __riscv_xtheadvector" + #endif + }] +} + + # Return 1 if we can execute code when using dg-add-options riscv_v proc check_effective_target_riscv_v_ok { } { -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re：Re：[PATCH v5] RISC-V: Handle differences between XTheadvector and Vector 2024-01-10 7:17 ` Re：[PATCH " juzhe.zhong 2024-01-10 7:26 ` 回复：Re：[PATCH " joshua @ 2024-01-10 7:28 ` joshua 1 sibling, 0 replies; 133+ messages in thread From: joshua @ 2024-01-10 7:28 UTC (permalink / raw) To: juzhe.zhong, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu I'm confused why I cannot add new shapes. I think adding new shapes is the basic part in implementation for new intrinsics. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月10日(星期三) 15:17 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Handle differences between XTheadvector and Vector Why do you need to invade existing shapes ? juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-10 15:16 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Handle differences between XTheadvector and Vector These xttheadvector speical intrinsics are different from rvv1.0 in determining function name from base name. We cannot directly reuse the existing shapes. In order not to invade existing shapes, we add new shapes for new functions. Also, we create new thead-vector-builtins.cc for xtheadvector function_base implementation. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月10日(星期三) 15:01 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Handle differences between XTheadvector and Vector Why do you add theadvector shapes ? I think you can reuse the current existing shapes. +thead-vector-builtins.o: \+ $(srcdir)/config/riscv/thead-vector-builtins.cc \+ $(CONFIG_H) $(SYSTEM_H) coretypes.h $(TM_H) $(TREE_H) $(RTL_H) \+ $(TM_P_H) memmodel.h insn-codes.h $(OPTABS_H) $(RECOG_H) \+ $(EXPR_H) $(BASIC_BLOCK_H) $(FUNCTION_H) fold-const.h $(GIMPLE_H) \+ gimple-iterator.h gimplify.h explow.h $(EMIT_RTL_H) tree-vector-builder.h \+ rtx-vector-builder.h \+ $(srcdir)/config/riscv/riscv-vector-builtins-shapes.h \+ $(srcdir)/config/riscv/riscv-vector-builtins-bases.h \+ $(srcdir)/config/riscv/thead-vector-builtins.h \+ $(RISCV_BUILTINS_H)+ $(COMPILER) -c $(ALL_COMPILERFLAGS) $(ALL_CPPFLAGS) $(INCLUDES) \+ $(srcdir)/config/riscv/thead-vector-builtins.cc+ Why do you rebuild another new object ? + Copyright (C) 2022-2023 Free Software Foundation, Inc. Incorrect copyright juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-10 10:57 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Handle differences between XTheadvector and Vector Hi Juzhe, Thank you for so many useful comments for this patch! There are some more patches to support xtheadvector special instrinsics as well as handle register overlap issue and rewrite assembly output. https://gcc.gnu.org/pipermail/gcc-patches/2024-January/641774.html https://gcc.gnu.org/pipermail/gcc-patches/2024-January/641732.html https://gcc.gnu.org/pipermail/gcc-patches/2024-January/641733.html Also, there is a precedent patch to refactor riscv-vector-builtins-bases.cc https://gcc.gnu.org/pipermail/gcc-patches/2023-December/641530.html Jeff has reviewed it, but didn't have LGTM yet. Joshua ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月10日(星期三) 10:34 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Handle differences between XTheadvector and Vector Thanks for your patience. LGTM from myside. I think it's pretty clean now. I can image in the future when some day the theadvector is no longer used, we can remove it very easily. And also, the theadvector won't affect our RVV1.0 maintain since it's isolated cleanly. But I'd like to wait for a few more days some body want to chime in. And you should do more things before commit it: 1. Remember you should run the full coverage RVV1.0 API test, the test-generator is downloaded from official intrinsic doc: https://github.com/riscv-non-isa/rvv-intrinsic-doc 2. Also the regression of RV32 an RV64 of GCC testsuite. Do you have more patches of theadvector that I didn't review ? plz point them to me again. Thanks. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-10 10:22 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Handle differences between XTheadvector and Vector This patch is to handle the differences in instruction generation between Vector and XTheadVector. In this version, we only support partial xtheadvector instructions that leverage directly from current RVV1.0 with simple adding "th." prefix. For different name xtheadvector instructions but share same patterns as RVV1.0 instructions, we will use ASM targethook to rewrite the whole string of the instructions in the following patches. For some vector patterns that cannot be avoided, we use "!TARGET_XTHEADVECTOR" to disable them in vector.md in order not to generate instructions that xtheadvector does not support, like vmv1r and vsext.vf2. gcc/ChangeLog: * config.gcc: Add files for XTheadVector intrinsics. * config/riscv/autovec.md: Guard XTheadVector. * config/riscv/riscv-c.cc: Add pragma for XTheadVector. * config/riscv/riscv-string.cc (expand_block_move): Guard XTheadVector. * config/riscv/riscv-string.cc (vls_mode_valid_p): Avoid autovec. * config/riscv/riscv-vector-builtins-shapes.cc (check_type): (build_one): New function. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): (DEF_THEAD_RVV_FUNCTION): Add new marcos. (check_required_extensions): (handle_pragma_vector): * config/riscv/riscv-vector-builtins.h (RVV_REQUIRE_VECTOR): (RVV_REQUIRE_XTHEADVECTOR): Add RVV_REQUIRE_VECTOR and RVV_REQUIRE_XTHEADVECTOR. (struct function_group_info): * config/riscv/riscv-vector-switch.def (ENTRY): Disable fractional mode for the XTheadVector extension. (TUPLE_ENTRY): Likewise. * config/riscv/riscv-vsetvl.cc: Add functions for xtheadvector. * config/riscv/riscv.cc (riscv_v_ext_vls_mode_p): Guard XTheadVector. (riscv_v_adjust_bytesize): Likewise. (riscv_preferred_simd_mode): Likewsie. (riscv_autovectorize_vector_modes): Likewise. (riscv_vector_mode_supported_any_target_p): Likewise. (TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P): Likewise. * config/riscv/vector.md: Include thead-vector.md. * config/riscv/riscv_th_vector.h: New file. * config/riscv/thead-vector.md: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/base/pragma-1.c: Add XTheadVector. * gcc.target/riscv/rvv/base/abi-1.c: Exclude XTheadVector. * lib/target-supports.exp: Add target for XTheadVector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config.gcc | 2 +- gcc/config/riscv/autovec.md | 2 +- gcc/config/riscv/predicates.md | 4 +- gcc/config/riscv/riscv-c.cc | 3 +- gcc/config/riscv/riscv-string.cc | 3 +- gcc/config/riscv/riscv-v.cc | 2 +- .../riscv/riscv-vector-builtins-bases.cc | 48 ++++-- .../riscv/riscv-vector-builtins-shapes.cc | 23 +++ gcc/config/riscv/riscv-vector-switch.def | 150 +++++++++--------- gcc/config/riscv/riscv.cc | 20 ++- gcc/config/riscv/riscv_th_vector.h | 49 ++++++ gcc/config/riscv/thead-vector.md | 102 ++++++++++++ gcc/config/riscv/thead.cc | 23 ++- gcc/config/riscv/vector.md | 49 ++++-- .../gcc.target/riscv/rvv/base/abi-1.c | 2 +- .../gcc.target/riscv/rvv/base/pragma-1.c | 2 +- gcc/testsuite/lib/target-supports.exp | 12 ++ 17 files changed, 383 insertions(+), 113 deletions(-) create mode 100644 gcc/config/riscv/riscv_th_vector.h create mode 100644 gcc/config/riscv/thead-vector.md diff --git a/gcc/config.gcc b/gcc/config.gcc index 7e583390024..047e4c02cf4 100644 --- a/gcc/config.gcc +++ b/gcc/config.gcc @@ -549,7 +549,7 @@ riscv*) extra_objs="${extra_objs} riscv-vector-builtins.o riscv-vector-builtins-shapes.o riscv-vector-builtins-bases.o" extra_objs="${extra_objs} thead.o riscv-target-attr.o" d_target_objs="riscv-d.o" - extra_headers="riscv_vector.h" + extra_headers="riscv_vector.h riscv_th_vector.h" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.cc" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.h" ;; diff --git a/gcc/config/riscv/autovec.md b/gcc/config/riscv/autovec.md index 775eaa825b0..0477781cabe 100644 --- a/gcc/config/riscv/autovec.md +++ b/gcc/config/riscv/autovec.md @@ -2579,7 +2579,7 @@ [(match_operand 0 "register_operand") (match_operand 1 "memory_operand") (match_operand:ANYI 2 "const_int_operand")] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" { riscv_vector::expand_rawmemchr(<MODE>mode, operands[0], operands[1], operands[2]); diff --git a/gcc/config/riscv/predicates.md b/gcc/config/riscv/predicates.md index b1a79cae50a..0337da88284 100644 --- a/gcc/config/riscv/predicates.md +++ b/gcc/config/riscv/predicates.md @@ -428,7 +428,9 @@ ;; Predicates for the V extension. (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (ior (match_test "TARGET_XTHEADVECTOR && rtx_equal_p (op, const0_rtx)") + (match_test "!TARGET_XTHEADVECTOR")) + (match_operand 0 "const_csr_operand")))) (define_special_predicate "autovec_length_operand" (ior (match_operand 0 "pmode_register_operand") diff --git a/gcc/config/riscv/riscv-c.cc b/gcc/config/riscv/riscv-c.cc index 422ddc2c308..2e306057347 100644 --- a/gcc/config/riscv/riscv-c.cc +++ b/gcc/config/riscv/riscv-c.cc @@ -195,7 +195,8 @@ riscv_pragma_intrinsic (cpp_reader *) const char *name = TREE_STRING_POINTER (x); - if (strcmp (name, "vector") == 0) + if (strcmp (name, "vector") == 0 + || strcmp (name, "xtheadvector") == 0) { if (!TARGET_VECTOR) { diff --git a/gcc/config/riscv/riscv-string.cc b/gcc/config/riscv/riscv-string.cc index f227b8667ce..b09b51d7526 100644 --- a/gcc/config/riscv/riscv-string.cc +++ b/gcc/config/riscv/riscv-string.cc @@ -773,7 +773,8 @@ riscv_expand_block_move_scalar (rtx dest, rtx src, rtx length) bool riscv_expand_block_move (rtx dest, rtx src, rtx length) { - if (TARGET_VECTOR && stringop_strategy & STRATEGY_VECTOR) + if ((TARGET_VECTOR && !TARGET_XTHEADVECTOR) + && stringop_strategy & STRATEGY_VECTOR) { bool ok = riscv_vector::expand_block_move (dest, src, length); if (ok) diff --git a/gcc/config/riscv/riscv-v.cc b/gcc/config/riscv/riscv-v.cc index 2491522191a..7e25deb966d 100644 --- a/gcc/config/riscv/riscv-v.cc +++ b/gcc/config/riscv/riscv-v.cc @@ -4392,7 +4392,7 @@ cmp_lmul_gt_one (machine_mode mode) bool vls_mode_valid_p (machine_mode vls_mode) { - if (!TARGET_VECTOR) + if (!TARGET_VECTOR || TARGET_XTHEADVECTOR) return false; if (riscv_autovec_preference == RVV_SCALABLE) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 91af2c00674..46f1a1da33e 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -115,23 +115,37 @@ public: tree type = builtin_types[e.type.index].vector; machine_mode mode = TYPE_MODE (type); - /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. - - - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 - - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 - - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 - - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 - - e8,m2/e16,m4/e32,m8 --> e8m2 - - e8,m4/e16,m8 --> e8m4 - - e8,m8 --> e8m8 - */ - /* SEW. */ - e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); - - /* LMUL. */ - machine_mode e8_mode - = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); - e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + + if (TARGET_XTHEADVECTOR) + { + machine_mode inner_mode = GET_MODE_INNER (mode); + /* SEW. */ + e.add_input_operand (Pmode, + gen_int_mode (GET_MODE_BITSIZE (inner_mode), Pmode)); + /* LMUL. */ + e.add_input_operand (Pmode, + gen_int_mode (get_vlmul (mode), Pmode)); + } + else + { + /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. + + - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 + - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 + - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 + - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 + - e8,m2/e16,m4/e32,m8 --> e8m2 + - e8,m4/e16,m8 --> e8m4 + - e8,m8 --> e8m8 + */ + /* SEW. */ + e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); + + /* LMUL. */ + machine_mode e8_mode + = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); + e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + } /* TAIL_ANY. */ e.add_input_operand (Pmode, diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index ee8058dcdb1..1e4f4d53de6 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -33,6 +33,25 @@ namespace riscv_vector { +/* Check whether the RETURN_TYPE and ARGUMENT_TYPES are + valid for the function. */ + +static bool +check_type (tree return_type, vec<tree> &argument_types) +{ + tree arg; + unsigned i; + + if (!return_type) + return false; + + FOR_EACH_VEC_ELT (argument_types, i, arg) + if (!arg) + return false; + + return true; +} + /* Add one function instance for GROUP, using operand suffix at index OI, mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ static void @@ -49,6 +68,10 @@ build_one (function_builder &b, const function_group_info &group, group.ops_infos.types[vec_type_idx].index); b.allocate_argument_types (function_instance, argument_types); b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + b.add_overloaded_function (function_instance, *group.shape); b.add_unique_function (function_instance, (*group.shape), return_type, argument_types); diff --git a/gcc/config/riscv/riscv-vector-switch.def b/gcc/config/riscv/riscv-vector-switch.def index 1ad26c2a3b2..452283b7416 100644 --- a/gcc/config/riscv/riscv-vector-switch.def +++ b/gcc/config/riscv/riscv-vector-switch.def @@ -68,9 +68,9 @@ Encode the ratio of SEW/LMUL into the mask types. #endif /* Disable modes if TARGET_MIN_VLEN == 32. */ -ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) -ENTRY (RVVMF32BI, true, LMUL_F4, 32) -ENTRY (RVVMF16BI, true, LMUL_F2, 16) +ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F8, 64) +ENTRY (RVVMF32BI, true, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F4, 32) +ENTRY (RVVMF16BI, true, TARGET_XTHEADVECTOR ? LMUL_1 : LMUL_F2 , 16) ENTRY (RVVMF8BI, true, LMUL_1, 8) ENTRY (RVVMF4BI, true, LMUL_2, 4) ENTRY (RVVMF2BI, true, LMUL_4, 2) @@ -81,39 +81,39 @@ ENTRY (RVVM8QI, true, LMUL_8, 1) ENTRY (RVVM4QI, true, LMUL_4, 2) ENTRY (RVVM2QI, true, LMUL_2, 4) ENTRY (RVVM1QI, true, LMUL_1, 8) -ENTRY (RVVMF2QI, true, LMUL_F2, 16) -ENTRY (RVVMF4QI, true, LMUL_F4, 32) -ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) +ENTRY (RVVMF2QI, !TARGET_XTHEADVECTOR, LMUL_F2, 16) +ENTRY (RVVMF4QI, !TARGET_XTHEADVECTOR, LMUL_F4, 32) +ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F8, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8HI, true, LMUL_8, 2) ENTRY (RVVM4HI, true, LMUL_4, 4) ENTRY (RVVM2HI, true, LMUL_2, 8) ENTRY (RVVM1HI, true, LMUL_1, 16) -ENTRY (RVVMF2HI, true, LMUL_F2, 32) -ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HI, !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_16. */ ENTRY (RVVM8HF, TARGET_VECTOR_ELEN_FP_16, LMUL_8, 2) ENTRY (RVVM4HF, TARGET_VECTOR_ELEN_FP_16, LMUL_4, 4) ENTRY (RVVM2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_2, 8) ENTRY (RVVM1HF, TARGET_VECTOR_ELEN_FP_16, LMUL_1, 16) -ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_F2, 32) -ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8SI, true, LMUL_8, 4) ENTRY (RVVM4SI, true, LMUL_4, 8) ENTRY (RVVM2SI, true, LMUL_2, 16) ENTRY (RVVM1SI, true, LMUL_1, 32) -ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_32. */ ENTRY (RVVM8SF, TARGET_VECTOR_ELEN_FP_32, LMUL_8, 4) ENTRY (RVVM4SF, TARGET_VECTOR_ELEN_FP_32, LMUL_4, 8) ENTRY (RVVM2SF, TARGET_VECTOR_ELEN_FP_32, LMUL_2, 16) ENTRY (RVVM1SF, TARGET_VECTOR_ELEN_FP_32, LMUL_1, 32) -ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if !TARGET_VECTOR_ELEN_64. */ ENTRY (RVVM8DI, TARGET_VECTOR_ELEN_64, LMUL_8, 8) @@ -140,127 +140,127 @@ ENTRY (RVVM1DF, TARGET_VECTOR_ELEN_FP_64, LMUL_1, 64) #endif TUPLE_ENTRY (RVVM1x8QI, true, RVVM1QI, 8, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x8QI, true, RVVMF2QI, 8, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x8QI, true, RVVMF4QI, 8, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 8, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x8QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 8, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x8QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 8, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 8, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x7QI, true, RVVM1QI, 7, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x7QI, true, RVVMF2QI, 7, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x7QI, true, RVVMF4QI, 7, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 7, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x7QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 7, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x7QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 7, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 7, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x6QI, true, RVVM1QI, 6, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x6QI, true, RVVMF2QI, 6, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x6QI, true, RVVMF4QI, 6, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 6, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x6QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 6, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x6QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 6, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 6, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x5QI, true, RVVM1QI, 5, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x5QI, true, RVVMF2QI, 5, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x5QI, true, RVVMF4QI, 5, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 5, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x5QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 5, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x5QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 5, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 5, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x4QI, true, RVVM2QI, 4, LMUL_2, 4) TUPLE_ENTRY (RVVM1x4QI, true, RVVM1QI, 4, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x4QI, true, RVVMF2QI, 4, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x4QI, true, RVVMF4QI, 4, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 4, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x4QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 4, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x4QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 4, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 4, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x3QI, true, RVVM2QI, 3, LMUL_2, 4) TUPLE_ENTRY (RVVM1x3QI, true, RVVM1QI, 3, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x3QI, true, RVVMF2QI, 3, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x3QI, true, RVVMF4QI, 3, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 3, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x3QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 3, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x3QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 3, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 3, LMUL_F8, 64) TUPLE_ENTRY (RVVM4x2QI, true, RVVM4QI, 2, LMUL_4, 2) TUPLE_ENTRY (RVVM2x2QI, true, RVVM2QI, 2, LMUL_2, 4) TUPLE_ENTRY (RVVM1x2QI, true, RVVM1QI, 2, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x2QI, true, RVVMF2QI, 2, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x2QI, true, RVVMF4QI, 2, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 2, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x2QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 2, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x2QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 2, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 2, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x8HI, true, RVVM1HI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HI, true, RVVMF2HI, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HI, true, RVVM1HI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HI, true, RVVMF2HI, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HI, true, RVVM1HI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HI, true, RVVMF2HI, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HI, true, RVVM1HI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HI, true, RVVMF2HI, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HI, true, RVVM2HI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HI, true, RVVM1HI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HI, true, RVVMF2HI, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HI, true, RVVM2HI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HI, true, RVVM1HI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HI, true, RVVMF2HI, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HI, true, RVVM4HI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HI, true, RVVM2HI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HI, true, RVVM1HI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HI, true, RVVMF2HI, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM4HF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8SI, true, RVVM1SI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SI, true, RVVM1SI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SI, true, RVVM1SI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SI, true, RVVM1SI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SI, true, RVVM2SI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SI, true, RVVM1SI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SI, true, RVVM2SI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SI, true, RVVM1SI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SI, true, RVVM4SI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SI, true, RVVM2SI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SI, true, RVVM1SI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM4SF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 8, LMUL_1, 16) TUPLE_ENTRY (RVVM1x7DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 7, LMUL_1, 16) diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index 8f74cd2e502..28cf1f414e9 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -1406,6 +1406,9 @@ riscv_v_adjust_bytesize (machine_mode mode, int scale) { if (riscv_v_ext_vector_mode_p (mode)) { + if (TARGET_XTHEADVECTOR) + return BYTES_PER_RISCV_VECTOR; + poly_int64 nunits = GET_MODE_NUNITS (mode); poly_int64 mode_size = GET_MODE_SIZE (mode); @@ -9995,7 +9998,7 @@ riscv_use_divmod_expander (void) static machine_mode riscv_preferred_simd_mode (scalar_mode mode) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::preferred_simd_mode (mode); return word_mode; @@ -10346,7 +10349,7 @@ riscv_mode_priority (int, int n) unsigned int riscv_autovectorize_vector_modes (vector_modes *modes, bool all) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::autovectorize_vector_modes (modes, all); return default_autovectorize_vector_modes (modes, all); @@ -10529,6 +10532,16 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) return false; } +/* Implements target hook vector_mode_supported_any_target_p. */ + +static bool +riscv_vector_mode_supported_any_target_p (machine_mode mode) +{ + if (TARGET_XTHEADVECTOR) + return false; + return true; +} + /* Initialize the GCC target structure. */ #undef TARGET_ASM_ALIGNED_HI_OP #define TARGET_ASM_ALIGNED_HI_OP "\t.half\t" @@ -10872,6 +10885,9 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) #undef TARGET_PREFERRED_ELSE_VALUE #define TARGET_PREFERRED_ELSE_VALUE riscv_preferred_else_value +#undef TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P +#define TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P riscv_vector_mode_supported_any_target_p + struct gcc_target targetm = TARGET_INITIALIZER; #include "gt-riscv.h" diff --git a/gcc/config/riscv/riscv_th_vector.h b/gcc/config/riscv/riscv_th_vector.h new file mode 100644 index 00000000000..b6b6738bdda --- /dev/null +++ b/gcc/config/riscv/riscv_th_vector.h @@ -0,0 +1,49 @@ +/* RISC-V 'XTheadVector' Extension intrinsics include file. + Copyright (C) 2024 Free Software Foundation, Inc. + + This file is part of GCC. + + GCC is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published + by the Free Software Foundation; either version 3, or (at your + option) any later version. + + GCC is distributed in the hope that it will be useful, but WITHOUT + ANY WARRANTY; without even the implied warranty of MERCHANTABILITY + or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public + License for more details. + + Under Section 7 of GPL version 3, you are granted additional + permissions described in the GCC Runtime Library Exception, version + 3.1, as published by the Free Software Foundation. + + You should have received a copy of the GNU General Public License and + a copy of the GCC Runtime Library Exception along with this program; + see the files COPYING3 and COPYING.RUNTIME respectively. If not, see + <" target="_blank">" target="_blank">" target="_blank">http://www.gnu.org/licenses/>. */ + +#ifndef __RISCV_TH_VECTOR_H +#define __RISCV_TH_VECTOR_H + +#include <stdint.h> +#include <stddef.h> + +#ifndef __riscv_xtheadvector +#error "XTheadVector intrinsics require the xtheadvector extension." +#else +#ifdef __cplusplus +extern "C" { +#endif + +/* NOTE: This implementation of riscv_th_vector.h is intentionally short. It does + not define the RVV types and intrinsic functions directly in C and C++ + code, but instead uses the following pragma to tell GCC to insert the + necessary type and function definitions itself. The net effect is the + same, and the file is a complete implementation of riscv_th_vector.h. */ +#pragma riscv intrinsic "xtheadvector" + +#ifdef __cplusplus +} +#endif // __cplusplus +#endif // __riscv_xtheadvector +#endif // __RISCV_TH_ECTOR_H diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md new file mode 100644 index 00000000000..696b815252d --- /dev/null +++ b/gcc/config/riscv/thead-vector.md @@ -0,0 +1,102 @@ +(define_c_enum "unspec" [ + UNSPEC_TH_VWLDST +]) + +(define_mode_iterator V_VLS_VT [V VLS VT]) +(define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) + +(define_split + [(set (match_operand:V_VB_VLS_VT 0 "reg_or_mem_operand") + (match_operand:V_VB_VLS_VT 1 "reg_or_mem_operand"))] + "TARGET_XTHEADVECTOR" + [(const_int 0)] + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + }) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:V_VLS_VT 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:V_VLS_VT + [(match_operand:V_VLS_VT 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2")]) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:VB 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:VB + [(match_operand:VB 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2") + (set (attr "sew") (const_int 8)) + (set (attr "vlmul") (symbol_ref "riscv_vector::LMUL_1"))]) + +(define_insn_and_split "*pred_th_mov<mode>" + [(set (match_operand:VB_VLS 0 "nonimmediate_operand" "=vr, m, vr, vr, vr") + (if_then_else:VB_VLS + (unspec:VB_VLS + [(match_operand:VB_VLS 1 "vector_all_trues_mask_operand" "Wc1, Wc1, Wc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) + (match_operand:VB_VLS 3 "vector_move_operand" " m, vr, vr, Wc0, Wc1") + (match_operand:VB_VLS 2 "vector_undef_operand" " vu, vu, vu, vu, vu")))] + "TARGET_XTHEADVECTOR" + "@ + # + # + vmcpy.m\t%0,%3 + vmclr.m\t%0 + vmset.m\t%0" + "&& !reload_completed" + [(const_int 0)] + { + if ((MEM_P (operands[0]) || MEM_P (operands[3])) + || (REG_P (operands[0]) && REG_P (operands[3]) + && INTVAL (operands[5]) == riscv_vector::VLMAX)) + { + emit_move_insn (operands[0], operands[3]); + DONE; + } + FAIL; + } + [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/config/riscv/thead.cc b/gcc/config/riscv/thead.cc index 00b2e14c6c0..cc04101b256 100644 --- a/gcc/config/riscv/thead.cc +++ b/gcc/config/riscv/thead.cc @@ -27,6 +27,7 @@ #include "backend.h" #include "tree.h" #include "rtl.h" +#include "insn-attr.h" #include "explow.h" #include "memmodel.h" #include "emit-rtl.h" @@ -890,8 +891,26 @@ th_asm_output_opcode (FILE *asm_out_file, const char *p) { /* We need to add th. prefix to all the xtheadvector instructions here.*/ - if (current_output_insn != NULL && p[0] == 'v') - fputs ("th.", asm_out_file); + if (current_output_insn != NULL) + { + if (get_attr_type (current_output_insn) == TYPE_VSETVL) + { + if (strstr (p, "zero")) + { + if (strstr (p, "zero,zero")) + return "th.vsetvli\tzero,zero,e%0,%m1"; + else + return "th.vsetvli\tzero,%0,e%1,%m2"; + } + else + { + return "th.vsetvli\t%0,%1,e%2,%m3"; + } + } + + if (p[0] == 'v') + fputs ("th.", asm_out_file); + } return p; } diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index 24b7b4394be..3eb6daafbc2 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -87,7 +87,7 @@ ;; check. However, we need default value of SEW for vsetvl instruction since there ;; is no field for ratio in the vsetvl instruction encoding. (define_attr "sew" "" - (cond [(eq_attr "mode" "RVVMF64BI,RVVMF32BI,RVVMF16BI,RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ + (cond [(eq_attr "mode" "RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ RVVM8QI,RVVM4QI,RVVM2QI,RVVM1QI,RVVMF2QI,RVVMF4QI,RVVMF8QI,\ RVVM1x8QI,RVVMF2x8QI,RVVMF4x8QI,RVVMF8x8QI,\ RVVM1x7QI,RVVMF2x7QI,RVVMF4x7QI,RVVMF8x7QI,\ @@ -99,6 +99,18 @@ V1QI,V2QI,V4QI,V8QI,V16QI,V32QI,V64QI,V128QI,V256QI,V512QI,V1024QI,V2048QI,V4096QI,\ V1BI,V2BI,V4BI,V8BI,V16BI,V32BI,V64BI,V128BI,V256BI,V512BI,V1024BI,V2048BI,V4096BI") (const_int 8) + (eq_attr "mode" "RVVMF16BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 16) + (const_int 8)) + (eq_attr "mode" "RVVMF32BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 32) + (const_int 8)) + (eq_attr "mode" "RVVMF64BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 64) + (const_int 8)) (eq_attr "mode" "RVVM8HI,RVVM4HI,RVVM2HI,RVVM1HI,RVVMF2HI,RVVMF4HI,\ RVVM1x8HI,RVVMF2x8HI,RVVMF4x8HI,\ RVVM1x7HI,RVVMF2x7HI,RVVMF4x7HI,\ @@ -159,9 +171,9 @@ (eq_attr "mode" "RVVM4QI,RVVMF2BI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2QI,RVVMF4BI") (symbol_ref "riscv_vector::LMUL_2") (eq_attr "mode" "RVVM1QI,RVVMF8BI") (symbol_ref "riscv_vector::LMUL_1") - (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "riscv_vector::LMUL_F2") - (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "riscv_vector::LMUL_F4") - (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "riscv_vector::LMUL_F8") + (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F2") + (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F4") + (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F8") (eq_attr "mode" "RVVM8HI") (symbol_ref "riscv_vector::LMUL_8") (eq_attr "mode" "RVVM4HI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2HI") (symbol_ref "riscv_vector::LMUL_2") @@ -436,6 +448,10 @@ vaeskf1,vaeskf2,vaesz,vsha2ms,vsha2ch,vsha2cl,vsm4k,vsm4r,\ vsm3me,vsm3c") (const_int INVALID_ATTRIBUTE) + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE) (eq_attr "mode" "RVVM8QI,RVVM1BI") (const_int 1) (eq_attr "mode" "RVVM4QI,RVVMF2BI") (const_int 2) (eq_attr "mode" "RVVM2QI,RVVMF4BI") (const_int 4) @@ -905,6 +921,8 @@ (symbol_ref "riscv_vector::FRM_DYN")] (symbol_ref "riscv_vector::FRM_NONE"))) +(include "thead-vector.md") + ;; ----------------------------------------------------------------- ;; ---- Miscellaneous Operations ;; ----------------------------------------------------------------- @@ -1078,6 +1096,12 @@ - We can not leave it to TARGET_SECONDARY_RELOAD since it happens before spilling. The clobber scratch is used by spilling fractional registers in IRA/LRA so it's too early. */ + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; @@ -1114,7 +1138,7 @@ (define_insn "*mov<mode>_whole" [(set (match_operand:V_WHOLE 0 "reg_or_mem_operand" "=vr, m,vr") (match_operand:V_WHOLE 1 "reg_or_mem_operand" " m,vr,vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "@ vl%m1re<sew>.v\t%0,%1 vs%m1r.v\t%1,%0 @@ -1135,6 +1159,13 @@ (match_operand:VB 1 "general_operand"))] "TARGET_VECTOR" { + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } + if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; }) @@ -1142,7 +1173,7 @@ (define_insn "*mov<mode>" [(set (match_operand:VB 0 "register_operand" "=vr") (match_operand:VB 1 "register_operand" " vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "vmv1r.v\t%0,%1" [(set_attr "type" "vmov") (set_attr "mode" "<MODE>")]) @@ -3692,7 +3723,7 @@ (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3713,7 +3744,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3734,7 +3765,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c index 2e0e12aa045..2eef9e1e1a8 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c @@ -1,4 +1,4 @@ -/* { dg-do compile } */ +/* { dg-do compile { target { ! riscv_xtheadvector } } } */ /* { dg-skip-if "test rvv intrinsic" { *-*-* } { "*" } { "-march=rv*v*" } } */ void foo0 () {__rvv_bool64_t t;} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c index 3d81b179235..ef329e30785 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c @@ -1,4 +1,4 @@ /* { dg-do compile } */ /* { dg-options "-O3 -march=rv32gc -mabi=ilp32d" } */ -#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' extension enabled} } */ +#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' or 'XTHEADVECTOR' extension enabled} } */ diff --git a/gcc/testsuite/lib/target-supports.exp b/gcc/testsuite/lib/target-supports.exp index b27c30b8c51..bf0ad5342d6 100644 --- a/gcc/testsuite/lib/target-supports.exp +++ b/gcc/testsuite/lib/target-supports.exp @@ -1956,6 +1956,18 @@ proc check_effective_target_riscv_zbb { } { }] } +# Return 1 if the target arch supports the XTheadVector extension, 0 otherwise. +# Cache the result. + +proc check_effective_target_riscv_xtheadvector { } { + return [check_no_compiler_messages riscv_ext_xtheadvector assembly { + #ifndef __riscv_xtheadvector + #error "Not __riscv_xtheadvector" + #endif + }] +} + + # Return 1 if we can execute code when using dg-add-options riscv_v proc check_effective_target_riscv_v_ok { } { -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* [PATCH v5] RISC-V: Handle differences between XTheadvector and Vector 2024-01-10 2:22 ` Jun Sha (Joshua) 2024-01-10 2:34 ` juzhe.zhong @ 2024-01-11 11:03 ` Jun Sha (Joshua) 1 sibling, 0 replies; 133+ messages in thread From: Jun Sha (Joshua) @ 2024-01-11 11:03 UTC (permalink / raw) To: gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, Jun Sha (Joshua), Jin Ma, Xianmiao Qu This patch is to handle the differences in instruction generation between Vector and XTheadVector. In this version, we only support partial xtheadvector instructions that leverage directly from current RVV1.0 with simple adding "th." prefix. For different name xtheadvector instructions but share same patterns as RVV1.0 instructions, we will use ASM targethook to rewrite the whole string of the instructions in the following patches. For some vector patterns that cannot be avoided, we use "!TARGET_XTHEADVECTOR" to disable them in vector.md in order not to generate instructions that xtheadvector does not support, like vmv1r. gcc/ChangeLog: * config.gcc: Add files for XTheadVector intrinsics. * config/riscv/autovec.md: Guard XTheadVector. * config/riscv/riscv-c.cc: Add pragma for XTheadVector. * config/riscv/riscv-string.cc (expand_block_move): Guard XTheadVector. * config/riscv/riscv-string.cc (vls_mode_valid_p): Avoid autovec. * config/riscv/riscv-vector-builtins-shapes.cc (check_type): (build_one): New function. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): (DEF_THEAD_RVV_FUNCTION): Add new marcos. (check_required_extensions): (handle_pragma_vector): * config/riscv/riscv-vector-builtins.h (RVV_REQUIRE_VECTOR): (RVV_REQUIRE_XTHEADVECTOR): Add RVV_REQUIRE_VECTOR and RVV_REQUIRE_XTHEADVECTOR. (struct function_group_info): * config/riscv/riscv-vector-switch.def (ENTRY): Disable fractional mode for the XTheadVector extension. (TUPLE_ENTRY): Likewise. * config/riscv/riscv-vsetvl.cc: Add functions for xtheadvector. * config/riscv/riscv.cc (riscv_v_ext_vls_mode_p): Guard XTheadVector. (riscv_v_adjust_bytesize): Likewise. (riscv_preferred_simd_mode): Likewsie. (riscv_autovectorize_vector_modes): Likewise. (riscv_vector_mode_supported_any_target_p): Likewise. (TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P): Likewise. * config/riscv/vector.md: Include thead-vector.md. * config/riscv/riscv_th_vector.h: New file. * config/riscv/thead-vector.md: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/base/pragma-1.c: Add XTheadVector. * gcc.target/riscv/rvv/base/abi-1.c: Exclude XTheadVector. * lib/target-supports.exp: Add target for XTheadVector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config.gcc | 2 +- gcc/config/riscv/autovec.md | 2 +- gcc/config/riscv/predicates.md | 4 +- gcc/config/riscv/riscv-c.cc | 3 +- gcc/config/riscv/riscv-string.cc | 3 +- gcc/config/riscv/riscv-v.cc | 2 +- .../riscv/riscv-vector-builtins-bases.cc | 48 ++++-- .../riscv/riscv-vector-builtins-shapes.cc | 23 +++ gcc/config/riscv/riscv-vector-switch.def | 150 +++++++++--------- gcc/config/riscv/riscv.cc | 20 ++- gcc/config/riscv/riscv_th_vector.h | 49 ++++++ gcc/config/riscv/thead-vector.md | 102 ++++++++++++ gcc/config/riscv/thead.cc | 23 ++- gcc/config/riscv/vector.md | 43 ++++- .../gcc.target/riscv/rvv/base/abi-1.c | 2 +- .../gcc.target/riscv/rvv/base/pragma-1.c | 2 +- gcc/testsuite/lib/target-supports.exp | 12 ++ 17 files changed, 380 insertions(+), 110 deletions(-) create mode 100644 gcc/config/riscv/riscv_th_vector.h create mode 100644 gcc/config/riscv/thead-vector.md diff --git a/gcc/config.gcc b/gcc/config.gcc index 7e583390024..047e4c02cf4 100644 --- a/gcc/config.gcc +++ b/gcc/config.gcc @@ -549,7 +549,7 @@ riscv*) extra_objs="${extra_objs} riscv-vector-builtins.o riscv-vector-builtins-shapes.o riscv-vector-builtins-bases.o" extra_objs="${extra_objs} thead.o riscv-target-attr.o" d_target_objs="riscv-d.o" - extra_headers="riscv_vector.h" + extra_headers="riscv_vector.h riscv_th_vector.h" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.cc" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.h" ;; diff --git a/gcc/config/riscv/autovec.md b/gcc/config/riscv/autovec.md index 775eaa825b0..0477781cabe 100644 --- a/gcc/config/riscv/autovec.md +++ b/gcc/config/riscv/autovec.md @@ -2579,7 +2579,7 @@ [(match_operand 0 "register_operand") (match_operand 1 "memory_operand") (match_operand:ANYI 2 "const_int_operand")] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" { riscv_vector::expand_rawmemchr(<MODE>mode, operands[0], operands[1], operands[2]); diff --git a/gcc/config/riscv/predicates.md b/gcc/config/riscv/predicates.md index b1a79cae50a..0337da88284 100644 --- a/gcc/config/riscv/predicates.md +++ b/gcc/config/riscv/predicates.md @@ -428,7 +428,9 @@ ;; Predicates for the V extension. (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (ior (match_test "TARGET_XTHEADVECTOR && rtx_equal_p (op, const0_rtx)") + (match_test "!TARGET_XTHEADVECTOR")) + (match_operand 0 "const_csr_operand")))) (define_special_predicate "autovec_length_operand" (ior (match_operand 0 "pmode_register_operand") diff --git a/gcc/config/riscv/riscv-c.cc b/gcc/config/riscv/riscv-c.cc index 422ddc2c308..2e306057347 100644 --- a/gcc/config/riscv/riscv-c.cc +++ b/gcc/config/riscv/riscv-c.cc @@ -195,7 +195,8 @@ riscv_pragma_intrinsic (cpp_reader *) const char *name = TREE_STRING_POINTER (x); - if (strcmp (name, "vector") == 0) + if (strcmp (name, "vector") == 0 + || strcmp (name, "xtheadvector") == 0) { if (!TARGET_VECTOR) { diff --git a/gcc/config/riscv/riscv-string.cc b/gcc/config/riscv/riscv-string.cc index f227b8667ce..b09b51d7526 100644 --- a/gcc/config/riscv/riscv-string.cc +++ b/gcc/config/riscv/riscv-string.cc @@ -773,7 +773,8 @@ riscv_expand_block_move_scalar (rtx dest, rtx src, rtx length) bool riscv_expand_block_move (rtx dest, rtx src, rtx length) { - if (TARGET_VECTOR && stringop_strategy & STRATEGY_VECTOR) + if ((TARGET_VECTOR && !TARGET_XTHEADVECTOR) + && stringop_strategy & STRATEGY_VECTOR) { bool ok = riscv_vector::expand_block_move (dest, src, length); if (ok) diff --git a/gcc/config/riscv/riscv-v.cc b/gcc/config/riscv/riscv-v.cc index 2491522191a..7e25deb966d 100644 --- a/gcc/config/riscv/riscv-v.cc +++ b/gcc/config/riscv/riscv-v.cc @@ -4392,7 +4392,7 @@ cmp_lmul_gt_one (machine_mode mode) bool vls_mode_valid_p (machine_mode vls_mode) { - if (!TARGET_VECTOR) + if (!TARGET_VECTOR || TARGET_XTHEADVECTOR) return false; if (riscv_autovec_preference == RVV_SCALABLE) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index fba98124473..1aa6e3c6665 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -131,23 +131,37 @@ public: tree type = builtin_types[e.type.index].vector; machine_mode mode = TYPE_MODE (type); - /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. - - - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 - - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 - - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 - - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 - - e8,m2/e16,m4/e32,m8 --> e8m2 - - e8,m4/e16,m8 --> e8m4 - - e8,m8 --> e8m8 - */ - /* SEW. */ - e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); - - /* LMUL. */ - machine_mode e8_mode - = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); - e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + + if (TARGET_XTHEADVECTOR) + { + machine_mode inner_mode = GET_MODE_INNER (mode); + /* SEW. */ + e.add_input_operand (Pmode, + gen_int_mode (GET_MODE_BITSIZE (inner_mode), Pmode)); + /* LMUL. */ + e.add_input_operand (Pmode, + gen_int_mode (get_vlmul (mode), Pmode)); + } + else + { + /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. + + - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 + - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 + - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 + - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 + - e8,m2/e16,m4/e32,m8 --> e8m2 + - e8,m4/e16,m8 --> e8m4 + - e8,m8 --> e8m8 + */ + /* SEW. */ + e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); + + /* LMUL. */ + machine_mode e8_mode + = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); + e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + } /* TAIL_ANY. */ e.add_input_operand (Pmode, diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index ee8058dcdb1..1e4f4d53de6 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -33,6 +33,25 @@ namespace riscv_vector { +/* Check whether the RETURN_TYPE and ARGUMENT_TYPES are + valid for the function. */ + +static bool +check_type (tree return_type, vec<tree> &argument_types) +{ + tree arg; + unsigned i; + + if (!return_type) + return false; + + FOR_EACH_VEC_ELT (argument_types, i, arg) + if (!arg) + return false; + + return true; +} + /* Add one function instance for GROUP, using operand suffix at index OI, mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ static void @@ -49,6 +68,10 @@ build_one (function_builder &b, const function_group_info &group, group.ops_infos.types[vec_type_idx].index); b.allocate_argument_types (function_instance, argument_types); b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + b.add_overloaded_function (function_instance, *group.shape); b.add_unique_function (function_instance, (*group.shape), return_type, argument_types); diff --git a/gcc/config/riscv/riscv-vector-switch.def b/gcc/config/riscv/riscv-vector-switch.def index 1ad26c2a3b2..452283b7416 100644 --- a/gcc/config/riscv/riscv-vector-switch.def +++ b/gcc/config/riscv/riscv-vector-switch.def @@ -68,9 +68,9 @@ Encode the ratio of SEW/LMUL into the mask types. #endif /* Disable modes if TARGET_MIN_VLEN == 32. */ -ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) -ENTRY (RVVMF32BI, true, LMUL_F4, 32) -ENTRY (RVVMF16BI, true, LMUL_F2, 16) +ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F8, 64) +ENTRY (RVVMF32BI, true, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F4, 32) +ENTRY (RVVMF16BI, true, TARGET_XTHEADVECTOR ? LMUL_1 : LMUL_F2 , 16) ENTRY (RVVMF8BI, true, LMUL_1, 8) ENTRY (RVVMF4BI, true, LMUL_2, 4) ENTRY (RVVMF2BI, true, LMUL_4, 2) @@ -81,39 +81,39 @@ ENTRY (RVVM8QI, true, LMUL_8, 1) ENTRY (RVVM4QI, true, LMUL_4, 2) ENTRY (RVVM2QI, true, LMUL_2, 4) ENTRY (RVVM1QI, true, LMUL_1, 8) -ENTRY (RVVMF2QI, true, LMUL_F2, 16) -ENTRY (RVVMF4QI, true, LMUL_F4, 32) -ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) +ENTRY (RVVMF2QI, !TARGET_XTHEADVECTOR, LMUL_F2, 16) +ENTRY (RVVMF4QI, !TARGET_XTHEADVECTOR, LMUL_F4, 32) +ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F8, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8HI, true, LMUL_8, 2) ENTRY (RVVM4HI, true, LMUL_4, 4) ENTRY (RVVM2HI, true, LMUL_2, 8) ENTRY (RVVM1HI, true, LMUL_1, 16) -ENTRY (RVVMF2HI, true, LMUL_F2, 32) -ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HI, !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_16. */ ENTRY (RVVM8HF, TARGET_VECTOR_ELEN_FP_16, LMUL_8, 2) ENTRY (RVVM4HF, TARGET_VECTOR_ELEN_FP_16, LMUL_4, 4) ENTRY (RVVM2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_2, 8) ENTRY (RVVM1HF, TARGET_VECTOR_ELEN_FP_16, LMUL_1, 16) -ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_F2, 32) -ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8SI, true, LMUL_8, 4) ENTRY (RVVM4SI, true, LMUL_4, 8) ENTRY (RVVM2SI, true, LMUL_2, 16) ENTRY (RVVM1SI, true, LMUL_1, 32) -ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_32. */ ENTRY (RVVM8SF, TARGET_VECTOR_ELEN_FP_32, LMUL_8, 4) ENTRY (RVVM4SF, TARGET_VECTOR_ELEN_FP_32, LMUL_4, 8) ENTRY (RVVM2SF, TARGET_VECTOR_ELEN_FP_32, LMUL_2, 16) ENTRY (RVVM1SF, TARGET_VECTOR_ELEN_FP_32, LMUL_1, 32) -ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if !TARGET_VECTOR_ELEN_64. */ ENTRY (RVVM8DI, TARGET_VECTOR_ELEN_64, LMUL_8, 8) @@ -140,127 +140,127 @@ ENTRY (RVVM1DF, TARGET_VECTOR_ELEN_FP_64, LMUL_1, 64) #endif TUPLE_ENTRY (RVVM1x8QI, true, RVVM1QI, 8, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x8QI, true, RVVMF2QI, 8, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x8QI, true, RVVMF4QI, 8, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 8, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x8QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 8, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x8QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 8, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 8, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x7QI, true, RVVM1QI, 7, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x7QI, true, RVVMF2QI, 7, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x7QI, true, RVVMF4QI, 7, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 7, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x7QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 7, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x7QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 7, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 7, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x6QI, true, RVVM1QI, 6, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x6QI, true, RVVMF2QI, 6, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x6QI, true, RVVMF4QI, 6, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 6, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x6QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 6, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x6QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 6, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 6, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x5QI, true, RVVM1QI, 5, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x5QI, true, RVVMF2QI, 5, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x5QI, true, RVVMF4QI, 5, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 5, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x5QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 5, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x5QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 5, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 5, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x4QI, true, RVVM2QI, 4, LMUL_2, 4) TUPLE_ENTRY (RVVM1x4QI, true, RVVM1QI, 4, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x4QI, true, RVVMF2QI, 4, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x4QI, true, RVVMF4QI, 4, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 4, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x4QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 4, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x4QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 4, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 4, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x3QI, true, RVVM2QI, 3, LMUL_2, 4) TUPLE_ENTRY (RVVM1x3QI, true, RVVM1QI, 3, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x3QI, true, RVVMF2QI, 3, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x3QI, true, RVVMF4QI, 3, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 3, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x3QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 3, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x3QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 3, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 3, LMUL_F8, 64) TUPLE_ENTRY (RVVM4x2QI, true, RVVM4QI, 2, LMUL_4, 2) TUPLE_ENTRY (RVVM2x2QI, true, RVVM2QI, 2, LMUL_2, 4) TUPLE_ENTRY (RVVM1x2QI, true, RVVM1QI, 2, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x2QI, true, RVVMF2QI, 2, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x2QI, true, RVVMF4QI, 2, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 2, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x2QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 2, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x2QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 2, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 2, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x8HI, true, RVVM1HI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HI, true, RVVMF2HI, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HI, true, RVVM1HI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HI, true, RVVMF2HI, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HI, true, RVVM1HI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HI, true, RVVMF2HI, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HI, true, RVVM1HI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HI, true, RVVMF2HI, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HI, true, RVVM2HI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HI, true, RVVM1HI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HI, true, RVVMF2HI, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HI, true, RVVM2HI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HI, true, RVVM1HI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HI, true, RVVMF2HI, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HI, true, RVVM4HI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HI, true, RVVM2HI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HI, true, RVVM1HI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HI, true, RVVMF2HI, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM4HF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8SI, true, RVVM1SI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SI, true, RVVM1SI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SI, true, RVVM1SI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SI, true, RVVM1SI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SI, true, RVVM2SI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SI, true, RVVM1SI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SI, true, RVVM2SI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SI, true, RVVM1SI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SI, true, RVVM4SI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SI, true, RVVM2SI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SI, true, RVVM1SI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM4SF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 8, LMUL_1, 16) TUPLE_ENTRY (RVVM1x7DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 7, LMUL_1, 16) diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index 8f74cd2e502..28cf1f414e9 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -1406,6 +1406,9 @@ riscv_v_adjust_bytesize (machine_mode mode, int scale) { if (riscv_v_ext_vector_mode_p (mode)) { + if (TARGET_XTHEADVECTOR) + return BYTES_PER_RISCV_VECTOR; + poly_int64 nunits = GET_MODE_NUNITS (mode); poly_int64 mode_size = GET_MODE_SIZE (mode); @@ -9995,7 +9998,7 @@ riscv_use_divmod_expander (void) static machine_mode riscv_preferred_simd_mode (scalar_mode mode) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::preferred_simd_mode (mode); return word_mode; @@ -10346,7 +10349,7 @@ riscv_mode_priority (int, int n) unsigned int riscv_autovectorize_vector_modes (vector_modes *modes, bool all) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::autovectorize_vector_modes (modes, all); return default_autovectorize_vector_modes (modes, all); @@ -10529,6 +10532,16 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) return false; } +/* Implements target hook vector_mode_supported_any_target_p. */ + +static bool +riscv_vector_mode_supported_any_target_p (machine_mode mode) +{ + if (TARGET_XTHEADVECTOR) + return false; + return true; +} + /* Initialize the GCC target structure. */ #undef TARGET_ASM_ALIGNED_HI_OP #define TARGET_ASM_ALIGNED_HI_OP "\t.half\t" @@ -10872,6 +10885,9 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) #undef TARGET_PREFERRED_ELSE_VALUE #define TARGET_PREFERRED_ELSE_VALUE riscv_preferred_else_value +#undef TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P +#define TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P riscv_vector_mode_supported_any_target_p + struct gcc_target targetm = TARGET_INITIALIZER; #include "gt-riscv.h" diff --git a/gcc/config/riscv/riscv_th_vector.h b/gcc/config/riscv/riscv_th_vector.h new file mode 100644 index 00000000000..b6b6738bdda --- /dev/null +++ b/gcc/config/riscv/riscv_th_vector.h @@ -0,0 +1,49 @@ +/* RISC-V 'XTheadVector' Extension intrinsics include file. + Copyright (C) 2024 Free Software Foundation, Inc. + + This file is part of GCC. + + GCC is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published + by the Free Software Foundation; either version 3, or (at your + option) any later version. + + GCC is distributed in the hope that it will be useful, but WITHOUT + ANY WARRANTY; without even the implied warranty of MERCHANTABILITY + or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public + License for more details. + + Under Section 7 of GPL version 3, you are granted additional + permissions described in the GCC Runtime Library Exception, version + 3.1, as published by the Free Software Foundation. + + You should have received a copy of the GNU General Public License and + a copy of the GCC Runtime Library Exception along with this program; + see the files COPYING3 and COPYING.RUNTIME respectively. If not, see + <http://www.gnu.org/licenses/>. */ + +#ifndef __RISCV_TH_VECTOR_H +#define __RISCV_TH_VECTOR_H + +#include <stdint.h> +#include <stddef.h> + +#ifndef __riscv_xtheadvector +#error "XTheadVector intrinsics require the xtheadvector extension." +#else +#ifdef __cplusplus +extern "C" { +#endif + +/* NOTE: This implementation of riscv_th_vector.h is intentionally short. It does + not define the RVV types and intrinsic functions directly in C and C++ + code, but instead uses the following pragma to tell GCC to insert the + necessary type and function definitions itself. The net effect is the + same, and the file is a complete implementation of riscv_th_vector.h. */ +#pragma riscv intrinsic "xtheadvector" + +#ifdef __cplusplus +} +#endif // __cplusplus +#endif // __riscv_xtheadvector +#endif // __RISCV_TH_ECTOR_H diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md new file mode 100644 index 00000000000..696b815252d --- /dev/null +++ b/gcc/config/riscv/thead-vector.md @@ -0,0 +1,102 @@ +(define_c_enum "unspec" [ + UNSPEC_TH_VWLDST +]) + +(define_mode_iterator V_VLS_VT [V VLS VT]) +(define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) + +(define_split + [(set (match_operand:V_VB_VLS_VT 0 "reg_or_mem_operand") + (match_operand:V_VB_VLS_VT 1 "reg_or_mem_operand"))] + "TARGET_XTHEADVECTOR" + [(const_int 0)] + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + }) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:V_VLS_VT 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:V_VLS_VT + [(match_operand:V_VLS_VT 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2")]) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:VB 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:VB + [(match_operand:VB 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2") + (set (attr "sew") (const_int 8)) + (set (attr "vlmul") (symbol_ref "riscv_vector::LMUL_1"))]) + +(define_insn_and_split "*pred_th_mov<mode>" + [(set (match_operand:VB_VLS 0 "nonimmediate_operand" "=vr, m, vr, vr, vr") + (if_then_else:VB_VLS + (unspec:VB_VLS + [(match_operand:VB_VLS 1 "vector_all_trues_mask_operand" "Wc1, Wc1, Wc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) + (match_operand:VB_VLS 3 "vector_move_operand" " m, vr, vr, Wc0, Wc1") + (match_operand:VB_VLS 2 "vector_undef_operand" " vu, vu, vu, vu, vu")))] + "TARGET_XTHEADVECTOR" + "@ + # + # + vmcpy.m\t%0,%3 + vmclr.m\t%0 + vmset.m\t%0" + "&& !reload_completed" + [(const_int 0)] + { + if ((MEM_P (operands[0]) || MEM_P (operands[3])) + || (REG_P (operands[0]) && REG_P (operands[3]) + && INTVAL (operands[5]) == riscv_vector::VLMAX)) + { + emit_move_insn (operands[0], operands[3]); + DONE; + } + FAIL; + } + [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/config/riscv/thead.cc b/gcc/config/riscv/thead.cc index 00b2e14c6c0..cc04101b256 100644 --- a/gcc/config/riscv/thead.cc +++ b/gcc/config/riscv/thead.cc @@ -27,6 +27,7 @@ #include "backend.h" #include "tree.h" #include "rtl.h" +#include "insn-attr.h" #include "explow.h" #include "memmodel.h" #include "emit-rtl.h" @@ -890,8 +891,26 @@ th_asm_output_opcode (FILE *asm_out_file, const char *p) { /* We need to add th. prefix to all the xtheadvector instructions here.*/ - if (current_output_insn != NULL && p[0] == 'v') - fputs ("th.", asm_out_file); + if (current_output_insn != NULL) + { + if (get_attr_type (current_output_insn) == TYPE_VSETVL) + { + if (strstr (p, "zero")) + { + if (strstr (p, "zero,zero")) + return "th.vsetvli\tzero,zero,e%0,%m1"; + else + return "th.vsetvli\tzero,%0,e%1,%m2"; + } + else + { + return "th.vsetvli\t%0,%1,e%2,%m3"; + } + } + + if (p[0] == 'v') + fputs ("th.", asm_out_file); + } return p; } diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index 24b7b4394be..34946090962 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -87,7 +87,7 @@ ;; check. However, we need default value of SEW for vsetvl instruction since there ;; is no field for ratio in the vsetvl instruction encoding. (define_attr "sew" "" - (cond [(eq_attr "mode" "RVVMF64BI,RVVMF32BI,RVVMF16BI,RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ + (cond [(eq_attr "mode" "RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ RVVM8QI,RVVM4QI,RVVM2QI,RVVM1QI,RVVMF2QI,RVVMF4QI,RVVMF8QI,\ RVVM1x8QI,RVVMF2x8QI,RVVMF4x8QI,RVVMF8x8QI,\ RVVM1x7QI,RVVMF2x7QI,RVVMF4x7QI,RVVMF8x7QI,\ @@ -99,6 +99,18 @@ V1QI,V2QI,V4QI,V8QI,V16QI,V32QI,V64QI,V128QI,V256QI,V512QI,V1024QI,V2048QI,V4096QI,\ V1BI,V2BI,V4BI,V8BI,V16BI,V32BI,V64BI,V128BI,V256BI,V512BI,V1024BI,V2048BI,V4096BI") (const_int 8) + (eq_attr "mode" "RVVMF16BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 16) + (const_int 8)) + (eq_attr "mode" "RVVMF32BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 32) + (const_int 8)) + (eq_attr "mode" "RVVMF64BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 64) + (const_int 8)) (eq_attr "mode" "RVVM8HI,RVVM4HI,RVVM2HI,RVVM1HI,RVVMF2HI,RVVMF4HI,\ RVVM1x8HI,RVVMF2x8HI,RVVMF4x8HI,\ RVVM1x7HI,RVVMF2x7HI,RVVMF4x7HI,\ @@ -159,9 +171,9 @@ (eq_attr "mode" "RVVM4QI,RVVMF2BI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2QI,RVVMF4BI") (symbol_ref "riscv_vector::LMUL_2") (eq_attr "mode" "RVVM1QI,RVVMF8BI") (symbol_ref "riscv_vector::LMUL_1") - (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "riscv_vector::LMUL_F2") - (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "riscv_vector::LMUL_F4") - (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "riscv_vector::LMUL_F8") + (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F2") + (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F4") + (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F8") (eq_attr "mode" "RVVM8HI") (symbol_ref "riscv_vector::LMUL_8") (eq_attr "mode" "RVVM4HI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2HI") (symbol_ref "riscv_vector::LMUL_2") @@ -436,6 +448,10 @@ vaeskf1,vaeskf2,vaesz,vsha2ms,vsha2ch,vsha2cl,vsm4k,vsm4r,\ vsm3me,vsm3c") (const_int INVALID_ATTRIBUTE) + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE) (eq_attr "mode" "RVVM8QI,RVVM1BI") (const_int 1) (eq_attr "mode" "RVVM4QI,RVVMF2BI") (const_int 2) (eq_attr "mode" "RVVM2QI,RVVMF4BI") (const_int 4) @@ -905,6 +921,8 @@ (symbol_ref "riscv_vector::FRM_DYN")] (symbol_ref "riscv_vector::FRM_NONE"))) +(include "thead-vector.md") + ;; ----------------------------------------------------------------- ;; ---- Miscellaneous Operations ;; ----------------------------------------------------------------- @@ -1078,6 +1096,12 @@ - We can not leave it to TARGET_SECONDARY_RELOAD since it happens before spilling. The clobber scratch is used by spilling fractional registers in IRA/LRA so it's too early. */ + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; @@ -1114,7 +1138,7 @@ (define_insn "*mov<mode>_whole" [(set (match_operand:V_WHOLE 0 "reg_or_mem_operand" "=vr, m,vr") (match_operand:V_WHOLE 1 "reg_or_mem_operand" " m,vr,vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "@ vl%m1re<sew>.v\t%0,%1 vs%m1r.v\t%1,%0 @@ -1135,6 +1159,13 @@ (match_operand:VB 1 "general_operand"))] "TARGET_VECTOR" { + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } + if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; }) @@ -1142,7 +1173,7 @@ (define_insn "*mov<mode>" [(set (match_operand:VB 0 "register_operand" "=vr") (match_operand:VB 1 "register_operand" " vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "vmv1r.v\t%0,%1" [(set_attr "type" "vmov") (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c index 2e0e12aa045..2eef9e1e1a8 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c @@ -1,4 +1,4 @@ -/* { dg-do compile } */ +/* { dg-do compile { target { ! riscv_xtheadvector } } } */ /* { dg-skip-if "test rvv intrinsic" { *-*-* } { "*" } { "-march=rv*v*" } } */ void foo0 () {__rvv_bool64_t t;} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c index 3d81b179235..ef329e30785 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c @@ -1,4 +1,4 @@ /* { dg-do compile } */ /* { dg-options "-O3 -march=rv32gc -mabi=ilp32d" } */ -#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' extension enabled} } */ +#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' or 'XTHEADVECTOR' extension enabled} } */ diff --git a/gcc/testsuite/lib/target-supports.exp b/gcc/testsuite/lib/target-supports.exp index b27c30b8c51..bf0ad5342d6 100644 --- a/gcc/testsuite/lib/target-supports.exp +++ b/gcc/testsuite/lib/target-supports.exp @@ -1956,6 +1956,18 @@ proc check_effective_target_riscv_zbb { } { }] } +# Return 1 if the target arch supports the XTheadVector extension, 0 otherwise. +# Cache the result. + +proc check_effective_target_riscv_xtheadvector { } { + return [check_no_compiler_messages riscv_ext_xtheadvector assembly { + #ifndef __riscv_xtheadvector + #error "Not __riscv_xtheadvector" + #endif + }] +} + + # Return 1 if we can execute code when using dg-add-options riscv_v proc check_effective_target_riscv_v_ok { } { -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: [PATCH v4] RISC-V: Handle differences between XTheadvector and Vector 2024-01-03 6:15 ` Jun Sha (Joshua) 2024-01-04 2:29 ` Jun Sha (Joshua) @ 2024-01-08 23:08 ` 钟居哲 2024-01-09 2:12 ` Re：[PATCH " joshua 2024-01-09 3:23 ` joshua 1 sibling, 2 replies; 133+ messages in thread From: 钟居哲 @ 2024-01-08 23:08 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, Jeff Law, Christoph Müllner, cooper.joshua, jinma, Cooper Qu [-- Attachment #1: Type: text/plain, Size: 69174 bytes --] - return TAIL_ANY; + return TARGET_XTHEADVECTOR ? TAIL_AGNOSTIC : TAIL_ANY; - return MASK_ANY; + return TARGET_XTHEADVECTOR ? MASK_UNDISTURBED : MASK_ANY; You shouldn't change this. - "vset%i1vli\t%0,%1,e%2,%m3,t%p4,m%p5" + { return TARGET_XTHEADVECTOR ? "vsetvli\t%0,%1,e%2,%m3" : "vset%i1vli\t%0,%1,e%2,%m3,t%p4,m%p5"; } I prefer do it in ASM_OUTPUT + Copyright (C) 2022-2023 Free Software Foundation, Inc. Copyright is not correct. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-03 14:15 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v4] RISC-V: Handle differences between XTheadvector and Vector This patch is to handle the differences in instruction generation between Vector and XTheadVector. In this version, we only support partial xtheadvector instructions that leverage directly from current RVV1.0 with simple adding "th." prefix. For different name xtheadvector instructions but share same patterns as RVV1.0 instructions, we will use ASM targethook to rewrite the whole string of the instructions in the following patches. For some vector patterns that cannot be avoided, we use "!TARGET_XTHEADVECTOR" to disable them in vector.md in order not to generate instructions that xtheadvector does not support, like vmv1r and vsext.vf2. gcc/ChangeLog: * config.gcc: Add files for XTheadVector intrinsics. * config/riscv/autovec.md: Guard XTheadVector. * config/riscv/riscv-c.cc: Add pragma for XTheadVector. * config/riscv/riscv-string.cc (expand_block_move): Guard XTheadVector. (get_prefer_tail_policy): Give specific value for tail. (get_prefer_mask_policy): Give specific value for mask. (vls_mode_valid_p): Avoid autovec. * config/riscv/riscv-vector-builtins-shapes.cc (check_type): (build_one): New function. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): (DEF_THEAD_RVV_FUNCTION): Add new marcos. (check_required_extensions): (handle_pragma_vector): * config/riscv/riscv-vector-builtins.h (RVV_REQUIRE_VECTOR): (RVV_REQUIRE_XTHEADVECTOR): Add RVV_REQUIRE_VECTOR and RVV_REQUIRE_XTHEADVECTOR. (struct function_group_info): * config/riscv/riscv-vector-switch.def (ENTRY): Disable fractional mode for the XTheadVector extension. (TUPLE_ENTRY): Likewise. * config/riscv/riscv-vsetvl.cc: Add functions for xtheadvector. * config/riscv/riscv.cc (riscv_v_ext_vls_mode_p): Guard XTheadVector. (riscv_v_adjust_bytesize): Likewise. (riscv_preferred_simd_mode): Likewsie. (riscv_autovectorize_vector_modes): Likewise. (riscv_vector_mode_supported_any_target_p): Likewise. (TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P): Likewise. * config/riscv/vector-iterators.md: Remove fractional LMUL. * config/riscv/vector.md: Include thead-vector.md. * config/riscv/riscv_th_vector.h: New file. * config/riscv/thead-vector.md: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/base/pragma-1.c: Add XTheadVector. * gcc.target/riscv/rvv/base/abi-1.c: Exclude XTheadVector. * lib/target-supports.exp: Add target for XTheadVector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config.gcc | 2 +- gcc/config/riscv/autovec.md | 2 +- gcc/config/riscv/predicates.md | 4 +- gcc/config/riscv/riscv-c.cc | 3 +- gcc/config/riscv/riscv-string.cc | 3 +- gcc/config/riscv/riscv-v.cc | 6 +- .../riscv/riscv-vector-builtins-bases.cc | 48 +++-- .../riscv/riscv-vector-builtins-shapes.cc | 23 +++ gcc/config/riscv/riscv-vector-switch.def | 150 +++++++------- gcc/config/riscv/riscv.cc | 20 +- gcc/config/riscv/riscv_th_vector.h | 49 +++++ gcc/config/riscv/thead-vector.md | 69 +++++++ gcc/config/riscv/vector-iterators.md | 186 +++++++++--------- gcc/config/riscv/vector.md | 55 ++++-- .../gcc.target/riscv/rvv/base/abi-1.c | 2 +- .../gcc.target/riscv/rvv/base/pragma-1.c | 2 +- gcc/testsuite/lib/target-supports.exp | 12 ++ 17 files changed, 427 insertions(+), 209 deletions(-) create mode 100644 gcc/config/riscv/riscv_th_vector.h create mode 100644 gcc/config/riscv/thead-vector.md diff --git a/gcc/config.gcc b/gcc/config.gcc index f0676c830e8..1445d98c147 100644 --- a/gcc/config.gcc +++ b/gcc/config.gcc @@ -549,7 +549,7 @@ riscv*) extra_objs="${extra_objs} riscv-vector-builtins.o riscv-vector-builtins-shapes.o riscv-vector-builtins-bases.o" extra_objs="${extra_objs} thead.o riscv-target-attr.o" d_target_objs="riscv-d.o" - extra_headers="riscv_vector.h" + extra_headers="riscv_vector.h riscv_th_vector.h" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.cc" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.h" ;; diff --git a/gcc/config/riscv/autovec.md b/gcc/config/riscv/autovec.md index 8b8a92f10a1..1fac56c7095 100644 --- a/gcc/config/riscv/autovec.md +++ b/gcc/config/riscv/autovec.md @@ -2579,7 +2579,7 @@ [(match_operand 0 "register_operand") (match_operand 1 "memory_operand") (match_operand:ANYI 2 "const_int_operand")] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" { riscv_vector::expand_rawmemchr(<MODE>mode, operands[0], operands[1], operands[2]); diff --git a/gcc/config/riscv/predicates.md b/gcc/config/riscv/predicates.md index b87a6900841..6e36c5afae1 100644 --- a/gcc/config/riscv/predicates.md +++ b/gcc/config/riscv/predicates.md @@ -428,7 +428,9 @@ ;; Predicates for the V extension. (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (ior (match_test "TARGET_XTHEADVECTOR && rtx_equal_p (op, const0_rtx)") + (match_test "!TARGET_XTHEADVECTOR")) + (match_operand 0 "const_csr_operand")))) (define_special_predicate "autovec_length_operand" (ior (match_operand 0 "pmode_register_operand") diff --git a/gcc/config/riscv/riscv-c.cc b/gcc/config/riscv/riscv-c.cc index d290c58613a..d6fb83ac00c 100644 --- a/gcc/config/riscv/riscv-c.cc +++ b/gcc/config/riscv/riscv-c.cc @@ -195,7 +195,8 @@ riscv_pragma_intrinsic (cpp_reader *) const char *name = TREE_STRING_POINTER (x); - if (strcmp (name, "vector") == 0) + if (strcmp (name, "vector") == 0 + || strcmp (name, "xtheadvector") == 0) { if (!TARGET_VECTOR) { diff --git a/gcc/config/riscv/riscv-string.cc b/gcc/config/riscv/riscv-string.cc index 11c1f74d0b3..f431b562222 100644 --- a/gcc/config/riscv/riscv-string.cc +++ b/gcc/config/riscv/riscv-string.cc @@ -773,7 +773,8 @@ riscv_expand_block_move_scalar (rtx dest, rtx src, rtx length) bool riscv_expand_block_move (rtx dest, rtx src, rtx length) { - if (TARGET_VECTOR && stringop_strategy & STRATEGY_VECTOR) + if ((TARGET_VECTOR && !TARGET_XTHEADVECTOR) + && stringop_strategy & STRATEGY_VECTOR) { bool ok = riscv_vector::expand_block_move (dest, src, length); if (ok) diff --git a/gcc/config/riscv/riscv-v.cc b/gcc/config/riscv/riscv-v.cc index 038ab084a37..eb061b593a5 100644 --- a/gcc/config/riscv/riscv-v.cc +++ b/gcc/config/riscv/riscv-v.cc @@ -1772,7 +1772,7 @@ get_prefer_tail_policy () compiler pick up either agnostic or undisturbed. Maybe we will have a compile option like -mprefer=agnostic to set this value???. */ - return TAIL_ANY; + return TARGET_XTHEADVECTOR ? TAIL_AGNOSTIC : TAIL_ANY; } /* Get prefer mask policy. */ @@ -1783,7 +1783,7 @@ get_prefer_mask_policy () compiler pick up either agnostic or undisturbed. Maybe we will have a compile option like -mprefer=agnostic to set this value???. */ - return MASK_ANY; + return TARGET_XTHEADVECTOR ? MASK_UNDISTURBED : MASK_ANY; } /* Get avl_type rtx. */ @@ -4383,7 +4383,7 @@ cmp_lmul_gt_one (machine_mode mode) bool vls_mode_valid_p (machine_mode vls_mode) { - if (!TARGET_VECTOR) + if (!TARGET_VECTOR || TARGET_XTHEADVECTOR) return false; if (riscv_autovec_preference == RVV_SCALABLE) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index c51affde353..22d877af9be 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -115,23 +115,37 @@ public: tree type = builtin_types[e.type.index].vector; machine_mode mode = TYPE_MODE (type); - /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. - - - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 - - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 - - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 - - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 - - e8,m2/e16,m4/e32,m8 --> e8m2 - - e8,m4/e16,m8 --> e8m4 - - e8,m8 --> e8m8 - */ - /* SEW. */ - e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); - - /* LMUL. */ - machine_mode e8_mode - = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); - e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + + if (TARGET_XTHEADVECTOR) + { + machine_mode inner_mode = GET_MODE_INNER (mode); + /* SEW. */ + e.add_input_operand (Pmode, + gen_int_mode (GET_MODE_BITSIZE (inner_mode), Pmode)); + /* LMUL. */ + e.add_input_operand (Pmode, + gen_int_mode (get_vlmul (mode), Pmode)); + } + else + { + /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. + + - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 + - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 + - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 + - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 + - e8,m2/e16,m4/e32,m8 --> e8m2 + - e8,m4/e16,m8 --> e8m4 + - e8,m8 --> e8m8 + */ + /* SEW. */ + e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); + + /* LMUL. */ + machine_mode e8_mode + = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); + e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + } /* TAIL_ANY. */ e.add_input_operand (Pmode, diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 4a754e0228f..6b49404a1fa 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -33,6 +33,25 @@ namespace riscv_vector { +/* Check whether the RETURN_TYPE and ARGUMENT_TYPES are + valid for the function. */ + +static bool +check_type (tree return_type, vec<tree> &argument_types) +{ + tree arg; + unsigned i; + + if (!return_type) + return false; + + FOR_EACH_VEC_ELT (argument_types, i, arg) + if (!arg) + return false; + + return true; +} + /* Add one function instance for GROUP, using operand suffix at index OI, mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ static void @@ -49,6 +68,10 @@ build_one (function_builder &b, const function_group_info &group, group.ops_infos.types[vec_type_idx].index); b.allocate_argument_types (function_instance, argument_types); b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + b.add_overloaded_function (function_instance, *group.shape); b.add_unique_function (function_instance, (*group.shape), return_type, argument_types); diff --git a/gcc/config/riscv/riscv-vector-switch.def b/gcc/config/riscv/riscv-vector-switch.def index 5c9f9bcbc3e..f7a66b34bae 100644 --- a/gcc/config/riscv/riscv-vector-switch.def +++ b/gcc/config/riscv/riscv-vector-switch.def @@ -68,9 +68,9 @@ Encode the ratio of SEW/LMUL into the mask types. #endif /* Disable modes if TARGET_MIN_VLEN == 32. */ -ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) -ENTRY (RVVMF32BI, true, LMUL_F4, 32) -ENTRY (RVVMF16BI, true, LMUL_F2, 16) +ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F8, 64) +ENTRY (RVVMF32BI, true, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F4, 32) +ENTRY (RVVMF16BI, true, TARGET_XTHEADVECTOR ? LMUL_1 : LMUL_F2 , 16) ENTRY (RVVMF8BI, true, LMUL_1, 8) ENTRY (RVVMF4BI, true, LMUL_2, 4) ENTRY (RVVMF2BI, true, LMUL_4, 2) @@ -81,39 +81,39 @@ ENTRY (RVVM8QI, true, LMUL_8, 1) ENTRY (RVVM4QI, true, LMUL_4, 2) ENTRY (RVVM2QI, true, LMUL_2, 4) ENTRY (RVVM1QI, true, LMUL_1, 8) -ENTRY (RVVMF2QI, true, LMUL_F2, 16) -ENTRY (RVVMF4QI, true, LMUL_F4, 32) -ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) +ENTRY (RVVMF2QI, !TARGET_XTHEADVECTOR, LMUL_F2, 16) +ENTRY (RVVMF4QI, !TARGET_XTHEADVECTOR, LMUL_F4, 32) +ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F8, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8HI, true, LMUL_8, 2) ENTRY (RVVM4HI, true, LMUL_4, 4) ENTRY (RVVM2HI, true, LMUL_2, 8) ENTRY (RVVM1HI, true, LMUL_1, 16) -ENTRY (RVVMF2HI, true, LMUL_F2, 32) -ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HI, !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_16. */ ENTRY (RVVM8HF, TARGET_VECTOR_ELEN_FP_16, LMUL_8, 2) ENTRY (RVVM4HF, TARGET_VECTOR_ELEN_FP_16, LMUL_4, 4) ENTRY (RVVM2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_2, 8) ENTRY (RVVM1HF, TARGET_VECTOR_ELEN_FP_16, LMUL_1, 16) -ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_F2, 32) -ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8SI, true, LMUL_8, 4) ENTRY (RVVM4SI, true, LMUL_4, 8) ENTRY (RVVM2SI, true, LMUL_2, 16) ENTRY (RVVM1SI, true, LMUL_1, 32) -ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_32. */ ENTRY (RVVM8SF, TARGET_VECTOR_ELEN_FP_32, LMUL_8, 4) ENTRY (RVVM4SF, TARGET_VECTOR_ELEN_FP_32, LMUL_4, 8) ENTRY (RVVM2SF, TARGET_VECTOR_ELEN_FP_32, LMUL_2, 16) ENTRY (RVVM1SF, TARGET_VECTOR_ELEN_FP_32, LMUL_1, 32) -ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if !TARGET_VECTOR_ELEN_64. */ ENTRY (RVVM8DI, TARGET_VECTOR_ELEN_64, LMUL_8, 8) @@ -140,127 +140,127 @@ ENTRY (RVVM1DF, TARGET_VECTOR_ELEN_FP_64, LMUL_1, 64) #endif TUPLE_ENTRY (RVVM1x8QI, true, RVVM1QI, 8, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x8QI, true, RVVMF2QI, 8, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x8QI, true, RVVMF4QI, 8, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 8, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x8QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 8, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x8QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 8, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 8, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x7QI, true, RVVM1QI, 7, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x7QI, true, RVVMF2QI, 7, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x7QI, true, RVVMF4QI, 7, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 7, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x7QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 7, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x7QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 7, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 7, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x6QI, true, RVVM1QI, 6, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x6QI, true, RVVMF2QI, 6, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x6QI, true, RVVMF4QI, 6, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 6, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x6QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 6, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x6QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 6, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 6, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x5QI, true, RVVM1QI, 5, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x5QI, true, RVVMF2QI, 5, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x5QI, true, RVVMF4QI, 5, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 5, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x5QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 5, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x5QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 5, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 5, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x4QI, true, RVVM2QI, 4, LMUL_2, 4) TUPLE_ENTRY (RVVM1x4QI, true, RVVM1QI, 4, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x4QI, true, RVVMF2QI, 4, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x4QI, true, RVVMF4QI, 4, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 4, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x4QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 4, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x4QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 4, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 4, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x3QI, true, RVVM2QI, 3, LMUL_2, 4) TUPLE_ENTRY (RVVM1x3QI, true, RVVM1QI, 3, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x3QI, true, RVVMF2QI, 3, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x3QI, true, RVVMF4QI, 3, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 3, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x3QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 3, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x3QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 3, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 3, LMUL_F8, 64) TUPLE_ENTRY (RVVM4x2QI, true, RVVM4QI, 2, LMUL_4, 2) TUPLE_ENTRY (RVVM2x2QI, true, RVVM2QI, 2, LMUL_2, 4) TUPLE_ENTRY (RVVM1x2QI, true, RVVM1QI, 2, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x2QI, true, RVVMF2QI, 2, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x2QI, true, RVVMF4QI, 2, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 2, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x2QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 2, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x2QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 2, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 2, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x8HI, true, RVVM1HI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HI, true, RVVMF2HI, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HI, true, RVVM1HI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HI, true, RVVMF2HI, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HI, true, RVVM1HI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HI, true, RVVMF2HI, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HI, true, RVVM1HI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HI, true, RVVMF2HI, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HI, true, RVVM2HI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HI, true, RVVM1HI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HI, true, RVVMF2HI, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HI, true, RVVM2HI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HI, true, RVVM1HI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HI, true, RVVMF2HI, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HI, true, RVVM4HI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HI, true, RVVM2HI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HI, true, RVVM1HI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HI, true, RVVMF2HI, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM4HF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8SI, true, RVVM1SI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SI, true, RVVM1SI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SI, true, RVVM1SI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SI, true, RVVM1SI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SI, true, RVVM2SI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SI, true, RVVM1SI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SI, true, RVVM2SI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SI, true, RVVM1SI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SI, true, RVVM4SI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SI, true, RVVM2SI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SI, true, RVVM1SI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM4SF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 8, LMUL_1, 16) TUPLE_ENTRY (RVVM1x7DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 7, LMUL_1, 16) diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index 51878797287..44a36db7803 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -1406,6 +1406,9 @@ riscv_v_adjust_bytesize (machine_mode mode, int scale) { if (riscv_v_ext_vector_mode_p (mode)) { + if (TARGET_XTHEADVECTOR) + return BYTES_PER_RISCV_VECTOR; + poly_int64 nunits = GET_MODE_NUNITS (mode); poly_int64 mode_size = GET_MODE_SIZE (mode); @@ -9962,7 +9965,7 @@ riscv_use_divmod_expander (void) static machine_mode riscv_preferred_simd_mode (scalar_mode mode) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::preferred_simd_mode (mode); return word_mode; @@ -10313,7 +10316,7 @@ riscv_mode_priority (int, int n) unsigned int riscv_autovectorize_vector_modes (vector_modes *modes, bool all) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::autovectorize_vector_modes (modes, all); return default_autovectorize_vector_modes (modes, all); @@ -10496,6 +10499,16 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) return false; } +/* Implements target hook vector_mode_supported_any_target_p. */ + +static bool +riscv_vector_mode_supported_any_target_p (machine_mode mode) +{ + if (TARGET_XTHEADVECTOR) + return false; + return true; +} + /* Initialize the GCC target structure. */ #undef TARGET_ASM_ALIGNED_HI_OP #define TARGET_ASM_ALIGNED_HI_OP "\t.half\t" @@ -10839,6 +10852,9 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) #undef TARGET_PREFERRED_ELSE_VALUE #define TARGET_PREFERRED_ELSE_VALUE riscv_preferred_else_value +#undef TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P +#define TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P riscv_vector_mode_supported_any_target_p + struct gcc_target targetm = TARGET_INITIALIZER; #include "gt-riscv.h" diff --git a/gcc/config/riscv/riscv_th_vector.h b/gcc/config/riscv/riscv_th_vector.h new file mode 100644 index 00000000000..a897b9cb90a --- /dev/null +++ b/gcc/config/riscv/riscv_th_vector.h @@ -0,0 +1,49 @@ +/* RISC-V 'XTheadVector' Extension intrinsics include file. + Copyright (C) 2022-2023 Free Software Foundation, Inc. + + This file is part of GCC. + + GCC is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published + by the Free Software Foundation; either version 3, or (at your + option) any later version. + + GCC is distributed in the hope that it will be useful, but WITHOUT + ANY WARRANTY; without even the implied warranty of MERCHANTABILITY + or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public + License for more details. + + Under Section 7 of GPL version 3, you are granted additional + permissions described in the GCC Runtime Library Exception, version + 3.1, as published by the Free Software Foundation. + + You should have received a copy of the GNU General Public License and + a copy of the GCC Runtime Library Exception along with this program; + see the files COPYING3 and COPYING.RUNTIME respectively. If not, see + <http://www.gnu.org/licenses/>. */ + +#ifndef __RISCV_TH_VECTOR_H +#define __RISCV_TH_VECTOR_H + +#include <stdint.h> +#include <stddef.h> + +#ifndef __riscv_xtheadvector +#error "XTheadVector intrinsics require the xtheadvector extension." +#else +#ifdef __cplusplus +extern "C" { +#endif + +/* NOTE: This implementation of riscv_th_vector.h is intentionally short. It does + not define the RVV types and intrinsic functions directly in C and C++ + code, but instead uses the following pragma to tell GCC to insert the + necessary type and function definitions itself. The net effect is the + same, and the file is a complete implementation of riscv_th_vector.h. */ +#pragma riscv intrinsic "xtheadvector" + +#ifdef __cplusplus +} +#endif // __cplusplus +#endif // __riscv_xtheadvector +#endif // __RISCV_TH_ECTOR_H diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md new file mode 100644 index 00000000000..1230bfb5994 --- /dev/null +++ b/gcc/config/riscv/thead-vector.md @@ -0,0 +1,69 @@ +(define_c_enum "unspec" [ + UNSPEC_TH_VWLDST +]) + +(define_mode_iterator V_VLS_VT [V VLS VT]) +(define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) + +(define_split + [(set (match_operand:V_VB_VLS_VT 0 "reg_or_mem_operand") + (match_operand:V_VB_VLS_VT 1 "reg_or_mem_operand"))] + "TARGET_XTHEADVECTOR" + [(const_int 0)] + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + }) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:V_VLS_VT 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:V_VLS_VT + [(match_operand:V_VLS_VT 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2")]) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:VB 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:VB + [(match_operand:VB 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2") + (set (attr "sew") (const_int 8)) + (set (attr "vlmul") (symbol_ref "riscv_vector::LMUL_1"))]) diff --git a/gcc/config/riscv/vector-iterators.md b/gcc/config/riscv/vector-iterators.md index 5f5f7b5b986..c0fc7a2441d 100644 --- a/gcc/config/riscv/vector-iterators.md +++ b/gcc/config/riscv/vector-iterators.md @@ -109,11 +109,11 @@ ]) (define_mode_iterator VI [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -128,11 +128,11 @@ ;; allow the instruction and mode to be matched during combine et al. (define_mode_iterator VF [ (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -140,11 +140,11 @@ (define_mode_iterator VF_ZVFHMIN [ (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -271,16 +271,16 @@ ]) (define_mode_iterator VEEWEXT2 [ - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -290,10 +290,10 @@ ]) (define_mode_iterator VEEWEXT4 [ - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -311,59 +311,59 @@ ]) (define_mode_iterator VEEWTRUNC2 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") (RVVM4SI "TARGET_64BIT") (RVVM2SI "TARGET_64BIT") (RVVM1SI "TARGET_64BIT") - (RVVMF2SI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEEWTRUNC4 [ - RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM2HI "TARGET_64BIT") (RVVM1HI "TARGET_64BIT") - (RVVMF2HI "TARGET_64BIT") - (RVVMF4HI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2HI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") (RVVM1HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEEWTRUNC8 [ (RVVM1QI "TARGET_64BIT") - (RVVMF2QI "TARGET_64BIT") - (RVVMF4QI "TARGET_64BIT") - (RVVMF8QI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2QI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF4QI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEI16 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -452,11 +452,11 @@ ]) (define_mode_iterator VFULLI [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_FULL_V") (RVVM4DI "TARGET_FULL_V") (RVVM2DI "TARGET_FULL_V") (RVVM1DI "TARGET_FULL_V") @@ -509,17 +509,17 @@ ]) (define_mode_iterator VI_QH [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") ]) (define_mode_iterator VI_QHS [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -560,11 +560,11 @@ ]) (define_mode_iterator VI_QHS_NO_M8 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -603,11 +603,11 @@ (define_mode_iterator VF_HS [ (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (V1HF "riscv_vector::vls_mode_valid_p (V1HFmode) && TARGET_ZVFH") (V2HF "riscv_vector::vls_mode_valid_p (V2HFmode) && TARGET_ZVFH") @@ -638,12 +638,12 @@ (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") - (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (V1HF "riscv_vector::vls_mode_valid_p (V1HFmode) && TARGET_ZVFH") (V2HF "riscv_vector::vls_mode_valid_p (V2HFmode) && TARGET_ZVFH") @@ -674,11 +674,11 @@ ]) (define_mode_iterator V_VLSI_QHS [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -756,27 +756,27 @@ ;; E.g. when index mode = RVVM8QImde and Pmode = SImode, if it is not zero_extend or ;; scalar != 1, such gather/scatter is not allowed since we don't have RVVM32SImode. (define_mode_iterator RATIO64 [ - (RVVMF8QI "TARGET_MIN_VLEN > 32") - (RVVMF4HI "TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM1DI "TARGET_VECTOR_ELEN_64") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") ]) (define_mode_iterator RATIO32 [ - RVVMF4QI - RVVMF2HI + (RVVMF4QI "!TARGET_XTHEADVECTOR") + (RVVMF2HI "!TARGET_XTHEADVECTOR") RVVM1SI (RVVM2DI "TARGET_VECTOR_ELEN_64") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") ]) (define_mode_iterator RATIO16 [ - RVVMF2QI + (RVVMF2QI "!TARGET_XTHEADVECTOR") RVVM1HI RVVM2SI (RVVM4DI "TARGET_VECTOR_ELEN_64") @@ -814,21 +814,21 @@ ]) (define_mode_iterator RATIO64I [ - (RVVMF8QI "TARGET_MIN_VLEN > 32") - (RVVMF4HI "TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM1DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") ]) (define_mode_iterator RATIO32I [ - RVVMF4QI - RVVMF2HI + (RVVMF4QI "!TARGET_XTHEADVECTOR") + (RVVMF2HI "!TARGET_XTHEADVECTOR") RVVM1SI (RVVM2DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") ]) (define_mode_iterator RATIO16I [ - RVVMF2QI + (RVVMF2QI "!TARGET_XTHEADVECTOR") RVVM1HI RVVM2SI (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") @@ -873,21 +873,21 @@ ]) (define_mode_iterator V_FRACT [ - RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") ]) (define_mode_iterator VWEXTI [ - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -933,7 +933,7 @@ (RVVM4SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -966,7 +966,7 @@ (RVVM4SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -996,7 +996,7 @@ (define_mode_iterator VWCONVERTI [ (RVVM8SI "TARGET_ZVFH") (RVVM4SI "TARGET_ZVFH") (RVVM2SI "TARGET_ZVFH") (RVVM1SI "TARGET_ZVFH") - (RVVMF2SI "TARGET_ZVFH") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_ZVFH") (RVVM8DI "TARGET_VECTOR_ELEN_64 && TARGET_VECTOR_ELEN_FP_32") (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_VECTOR_ELEN_FP_32") @@ -1045,7 +1045,7 @@ ]) (define_mode_iterator VQEXTI [ - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -1456,11 +1456,11 @@ ;; VINDEXED [VI8 VI16 VI32 (VI64 "TARGET_64BIT")]. (define_mode_iterator VINDEXED [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") @@ -1468,12 +1468,12 @@ (RVVM1DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64 && TARGET_64BIT") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64 && TARGET_64BIT") @@ -3173,11 +3173,11 @@ (define_mode_iterator V_VLS_F_CONVERT_SI [ (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") - (RVVMF2HF "TARGET_ZVFH") (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") @@ -3290,12 +3290,12 @@ ]) (define_mode_iterator V_VLS_F_CONVERT_DI [ - (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index b5a9055cdc4..cb30c9ae97c 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -83,7 +83,7 @@ ;; check. However, we need default value of SEW for vsetvl instruction since there ;; is no field for ratio in the vsetvl instruction encoding. (define_attr "sew" "" - (cond [(eq_attr "mode" "RVVMF64BI,RVVMF32BI,RVVMF16BI,RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ + (cond [(eq_attr "mode" "RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ RVVM8QI,RVVM4QI,RVVM2QI,RVVM1QI,RVVMF2QI,RVVMF4QI,RVVMF8QI,\ RVVM1x8QI,RVVMF2x8QI,RVVMF4x8QI,RVVMF8x8QI,\ RVVM1x7QI,RVVMF2x7QI,RVVMF4x7QI,RVVMF8x7QI,\ @@ -95,6 +95,18 @@ V1QI,V2QI,V4QI,V8QI,V16QI,V32QI,V64QI,V128QI,V256QI,V512QI,V1024QI,V2048QI,V4096QI,\ V1BI,V2BI,V4BI,V8BI,V16BI,V32BI,V64BI,V128BI,V256BI,V512BI,V1024BI,V2048BI,V4096BI") (const_int 8) + (eq_attr "mode" "RVVMF16BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 16) + (const_int 8)) + (eq_attr "mode" "RVVMF32BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 32) + (const_int 8)) + (eq_attr "mode" "RVVMF64BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 64) + (const_int 8)) (eq_attr "mode" "RVVM8HI,RVVM4HI,RVVM2HI,RVVM1HI,RVVMF2HI,RVVMF4HI,\ RVVM1x8HI,RVVMF2x8HI,RVVMF4x8HI,\ RVVM1x7HI,RVVMF2x7HI,RVVMF4x7HI,\ @@ -155,9 +167,9 @@ (eq_attr "mode" "RVVM4QI,RVVMF2BI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2QI,RVVMF4BI") (symbol_ref "riscv_vector::LMUL_2") (eq_attr "mode" "RVVM1QI,RVVMF8BI") (symbol_ref "riscv_vector::LMUL_1") - (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "riscv_vector::LMUL_F2") - (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "riscv_vector::LMUL_F4") - (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "riscv_vector::LMUL_F8") + (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F2") + (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F4") + (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F8") (eq_attr "mode" "RVVM8HI") (symbol_ref "riscv_vector::LMUL_8") (eq_attr "mode" "RVVM4HI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2HI") (symbol_ref "riscv_vector::LMUL_2") @@ -428,6 +440,10 @@ vislide1up,vislide1down,vfslide1up,vfslide1down,\ vgather,vcompress,vlsegdux,vlsegdox,vssegtux,vssegtox") (const_int INVALID_ATTRIBUTE) + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE) (eq_attr "mode" "RVVM8QI,RVVM1BI") (const_int 1) (eq_attr "mode" "RVVM4QI,RVVMF2BI") (const_int 2) (eq_attr "mode" "RVVM2QI,RVVMF4BI") (const_int 4) @@ -888,6 +904,8 @@ (symbol_ref "riscv_vector::FRM_DYN")] (symbol_ref "riscv_vector::FRM_NONE"))) +(include "thead-vector.md") + ;; ----------------------------------------------------------------- ;; ---- Miscellaneous Operations ;; ----------------------------------------------------------------- @@ -1061,6 +1079,12 @@ - We can not leave it to TARGET_SECONDARY_RELOAD since it happens before spilling. The clobber scratch is used by spilling fractional registers in IRA/LRA so it's too early. */ + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; @@ -1097,7 +1121,7 @@ (define_insn "*mov<mode>_whole" [(set (match_operand:V_WHOLE 0 "reg_or_mem_operand" "=vr, m,vr") (match_operand:V_WHOLE 1 "reg_or_mem_operand" " m,vr,vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "@ vl%m1re<sew>.v\t%0,%1 vs%m1r.v\t%1,%0 @@ -1118,6 +1142,13 @@ (match_operand:VB 1 "general_operand"))] "TARGET_VECTOR" { + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } + if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; }) @@ -1125,7 +1156,7 @@ (define_insn "*mov<mode>" [(set (match_operand:VB 0 "register_operand" "=vr") (match_operand:VB 1 "register_operand" " vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "vmv1r.v\t%0,%1" [(set_attr "type" "vmov") (set_attr "mode" "<MODE>")]) @@ -1511,7 +1542,7 @@ (match_dup 4) (match_dup 5)] UNSPEC_VSETVL))] "TARGET_VECTOR" - "vset%i1vli\t%0,%1,e%2,%m3,t%p4,m%p5" + { return TARGET_XTHEADVECTOR ? "vsetvli\t%0,%1,e%2,%m3" : "vset%i1vli\t%0,%1,e%2,%m3,t%p4,m%p5"; } [(set_attr "type" "vsetvl") (set_attr "mode" "<MODE>") (set (attr "sew") (symbol_ref "INTVAL (operands[2])")) @@ -1529,7 +1560,7 @@ (match_operand 2 "const_int_operand" "i") (match_operand 3 "const_int_operand" "i")] UNSPEC_VSETVL))] "TARGET_VECTOR" - "vsetvli\tzero,zero,e%0,%m1,t%p2,m%p3" + { return TARGET_XTHEADVECTOR ? "vsetvli\tzero,zero,e%0,%m1" : "vsetvli\tzero,zero,e%0,%m1,t%p2,m%p3"; } [(set_attr "type" "vsetvl") (set_attr "mode" "SI") (set (attr "sew") (symbol_ref "INTVAL (operands[0])")) @@ -1551,7 +1582,7 @@ (match_operand 3 "const_int_operand" "i") (match_operand 4 "const_int_operand" "i")] UNSPEC_VSETVL))] "TARGET_VECTOR" - "vset%i0vli\tzero,%0,e%1,%m2,t%p3,m%p4" + { return TARGET_XTHEADVECTOR ? "vsetvli\tzero,%0,e%1,%m2" : "vset%i0vli\tzero,%0,e%1,%m2,t%p3,m%p4"; } [(set_attr "type" "vsetvl") (set_attr "mode" "<MODE>") (set (attr "sew") (symbol_ref "INTVAL (operands[1])")) @@ -3680,7 +3711,7 @@ (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3701,7 +3732,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3722,7 +3753,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c index 2e0e12aa045..2eef9e1e1a8 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c @@ -1,4 +1,4 @@ -/* { dg-do compile } */ +/* { dg-do compile { target { ! riscv_xtheadvector } } } */ /* { dg-skip-if "test rvv intrinsic" { *-*-* } { "*" } { "-march=rv*v*" } } */ void foo0 () {__rvv_bool64_t t;} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c index 3d81b179235..ef329e30785 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c @@ -1,4 +1,4 @@ /* { dg-do compile } */ /* { dg-options "-O3 -march=rv32gc -mabi=ilp32d" } */ -#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' extension enabled} } */ +#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' or 'XTHEADVECTOR' extension enabled} } */ diff --git a/gcc/testsuite/lib/target-supports.exp b/gcc/testsuite/lib/target-supports.exp index 7f13ff0ca56..70df6b1401c 100644 --- a/gcc/testsuite/lib/target-supports.exp +++ b/gcc/testsuite/lib/target-supports.exp @@ -1952,6 +1952,18 @@ proc check_effective_target_riscv_zbb { } { }] } +# Return 1 if the target arch supports the XTheadVector extension, 0 otherwise. +# Cache the result. + +proc check_effective_target_riscv_xtheadvector { } { + return [check_no_compiler_messages riscv_ext_xtheadvector assembly { + #ifndef __riscv_xtheadvector + #error "Not __riscv_xtheadvector" + #endif + }] +} + + # Return 1 if we can execute code when using dg-add-options riscv_v proc check_effective_target_riscv_v_ok { } { -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re：[PATCH v4] RISC-V: Handle differences between XTheadvector and Vector 2024-01-08 23:08 ` [PATCH v4] " 钟居哲 @ 2024-01-09 2:12 ` joshua 2024-01-09 3:23 ` joshua 1 sibling, 0 replies; 133+ messages in thread From: joshua @ 2024-01-09 2:12 UTC (permalink / raw) To: 钟居哲, gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, Jeff Law, Christoph Müllner, jinma, Cooper Qu For the vsetvl issue, we have discussed last week. Maybe riscv_asm_output function cannot return instructions like riscv_output_move. The briefest approach may be to add some logic in the vsetvl patterns. Only 3 patterns need to be modified and that will not be too invasive. ------------------------------------------------------------------ 发件人：钟居哲 <juzhe.zhong@rivai.ai> 发送时间：2024年1月9日(星期二) 07:08 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送："jim.wilson.gcc"<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; Jeff Law<jeffreyalaw@gmail.com>; "Christoph Müllner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; Cooper Qu<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v4] RISC-V: Handle differences between XTheadvector and Vector - return TAIL_ANY; + return TARGET_XTHEADVECTOR ? TAIL_AGNOSTIC : TAIL_ANY; - return MASK_ANY; + return TARGET_XTHEADVECTOR ? MASK_UNDISTURBED : MASK_ANY; You shouldn't change this. - "vset%i1vli\t%0,%1,e%2,%m3,t%p4,m%p5" + { return TARGET_XTHEADVECTOR ? "vsetvli\t%0,%1,e%2,%m3" : "vset%i1vli\t%0,%1,e%2,%m3,t%p4,m%p5"; } I prefer do it in ASM_OUTPUT + Copyright (C) 2022-2023 Free Software Foundation, Inc. Copyright is not correct. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-03 14:15 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v4] RISC-V: Handle differences between XTheadvector and Vector This patch is to handle the differences in instruction generation between Vector and XTheadVector. In this version, we only support partial xtheadvector instructions that leverage directly from current RVV1.0 with simple adding "th." prefix. For different name xtheadvector instructions but share same patterns as RVV1.0 instructions, we will use ASM targethook to rewrite the whole string of the instructions in the following patches. For some vector patterns that cannot be avoided, we use "!TARGET_XTHEADVECTOR" to disable them in vector.md in order not to generate instructions that xtheadvector does not support, like vmv1r and vsext.vf2. gcc/ChangeLog: * config.gcc: Add files for XTheadVector intrinsics. * config/riscv/autovec.md: Guard XTheadVector. * config/riscv/riscv-c.cc: Add pragma for XTheadVector. * config/riscv/riscv-string.cc (expand_block_move): Guard XTheadVector. (get_prefer_tail_policy): Give specific value for tail. (get_prefer_mask_policy): Give specific value for mask. (vls_mode_valid_p): Avoid autovec. * config/riscv/riscv-vector-builtins-shapes.cc (check_type): (build_one): New function. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): (DEF_THEAD_RVV_FUNCTION): Add new marcos. (check_required_extensions): (handle_pragma_vector): * config/riscv/riscv-vector-builtins.h (RVV_REQUIRE_VECTOR): (RVV_REQUIRE_XTHEADVECTOR): Add RVV_REQUIRE_VECTOR and RVV_REQUIRE_XTHEADVECTOR. (struct function_group_info): * config/riscv/riscv-vector-switch.def (ENTRY): Disable fractional mode for the XTheadVector extension. (TUPLE_ENTRY): Likewise. * config/riscv/riscv-vsetvl.cc: Add functions for xtheadvector. * config/riscv/riscv.cc (riscv_v_ext_vls_mode_p): Guard XTheadVector. (riscv_v_adjust_bytesize): Likewise. (riscv_preferred_simd_mode): Likewsie. (riscv_autovectorize_vector_modes): Likewise. (riscv_vector_mode_supported_any_target_p): Likewise. (TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P): Likewise. * config/riscv/vector-iterators.md: Remove fractional LMUL. * config/riscv/vector.md: Include thead-vector.md. * config/riscv/riscv_th_vector.h: New file. * config/riscv/thead-vector.md: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/base/pragma-1.c: Add XTheadVector. * gcc.target/riscv/rvv/base/abi-1.c: Exclude XTheadVector. * lib/target-supports.exp: Add target for XTheadVector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config.gcc | 2 +- gcc/config/riscv/autovec.md | 2 +- gcc/config/riscv/predicates.md | 4 +- gcc/config/riscv/riscv-c.cc | 3 +- gcc/config/riscv/riscv-string.cc | 3 +- gcc/config/riscv/riscv-v.cc | 6 +- .../riscv/riscv-vector-builtins-bases.cc | 48 +++-- .../riscv/riscv-vector-builtins-shapes.cc | 23 +++ gcc/config/riscv/riscv-vector-switch.def | 150 +++++++------- gcc/config/riscv/riscv.cc | 20 +- gcc/config/riscv/riscv_th_vector.h | 49 +++++ gcc/config/riscv/thead-vector.md | 69 +++++++ gcc/config/riscv/vector-iterators.md | 186 +++++++++--------- gcc/config/riscv/vector.md | 55 ++++-- .../gcc.target/riscv/rvv/base/abi-1.c | 2 +- .../gcc.target/riscv/rvv/base/pragma-1.c | 2 +- gcc/testsuite/lib/target-supports.exp | 12 ++ 17 files changed, 427 insertions(+), 209 deletions(-) create mode 100644 gcc/config/riscv/riscv_th_vector.h create mode 100644 gcc/config/riscv/thead-vector.md diff --git a/gcc/config.gcc b/gcc/config.gcc index f0676c830e8..1445d98c147 100644 --- a/gcc/config.gcc +++ b/gcc/config.gcc @@ -549,7 +549,7 @@ riscv*) extra_objs="${extra_objs} riscv-vector-builtins.o riscv-vector-builtins-shapes.o riscv-vector-builtins-bases.o" extra_objs="${extra_objs} thead.o riscv-target-attr.o" d_target_objs="riscv-d.o" - extra_headers="riscv_vector.h" + extra_headers="riscv_vector.h riscv_th_vector.h" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.cc" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.h" ;; diff --git a/gcc/config/riscv/autovec.md b/gcc/config/riscv/autovec.md index 8b8a92f10a1..1fac56c7095 100644 --- a/gcc/config/riscv/autovec.md +++ b/gcc/config/riscv/autovec.md @@ -2579,7 +2579,7 @@ [(match_operand 0 "register_operand") (match_operand 1 "memory_operand") (match_operand:ANYI 2 "const_int_operand")] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" { riscv_vector::expand_rawmemchr(<MODE>mode, operands[0], operands[1], operands[2]); diff --git a/gcc/config/riscv/predicates.md b/gcc/config/riscv/predicates.md index b87a6900841..6e36c5afae1 100644 --- a/gcc/config/riscv/predicates.md +++ b/gcc/config/riscv/predicates.md @@ -428,7 +428,9 @@ ;; Predicates for the V extension. (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (ior (match_test "TARGET_XTHEADVECTOR && rtx_equal_p (op, const0_rtx)") + (match_test "!TARGET_XTHEADVECTOR")) + (match_operand 0 "const_csr_operand")))) (define_special_predicate "autovec_length_operand" (ior (match_operand 0 "pmode_register_operand") diff --git a/gcc/config/riscv/riscv-c.cc b/gcc/config/riscv/riscv-c.cc index d290c58613a..d6fb83ac00c 100644 --- a/gcc/config/riscv/riscv-c.cc +++ b/gcc/config/riscv/riscv-c.cc @@ -195,7 +195,8 @@ riscv_pragma_intrinsic (cpp_reader *) const char *name = TREE_STRING_POINTER (x); - if (strcmp (name, "vector") == 0) + if (strcmp (name, "vector") == 0 + || strcmp (name, "xtheadvector") == 0) { if (!TARGET_VECTOR) { diff --git a/gcc/config/riscv/riscv-string.cc b/gcc/config/riscv/riscv-string.cc index 11c1f74d0b3..f431b562222 100644 --- a/gcc/config/riscv/riscv-string.cc +++ b/gcc/config/riscv/riscv-string.cc @@ -773,7 +773,8 @@ riscv_expand_block_move_scalar (rtx dest, rtx src, rtx length) bool riscv_expand_block_move (rtx dest, rtx src, rtx length) { - if (TARGET_VECTOR && stringop_strategy & STRATEGY_VECTOR) + if ((TARGET_VECTOR && !TARGET_XTHEADVECTOR) + && stringop_strategy & STRATEGY_VECTOR) { bool ok = riscv_vector::expand_block_move (dest, src, length); if (ok) diff --git a/gcc/config/riscv/riscv-v.cc b/gcc/config/riscv/riscv-v.cc index 038ab084a37..eb061b593a5 100644 --- a/gcc/config/riscv/riscv-v.cc +++ b/gcc/config/riscv/riscv-v.cc @@ -1772,7 +1772,7 @@ get_prefer_tail_policy () compiler pick up either agnostic or undisturbed. Maybe we will have a compile option like -mprefer=agnostic to set this value???. */ - return TAIL_ANY; + return TARGET_XTHEADVECTOR ? TAIL_AGNOSTIC : TAIL_ANY; } /* Get prefer mask policy. */ @@ -1783,7 +1783,7 @@ get_prefer_mask_policy () compiler pick up either agnostic or undisturbed. Maybe we will have a compile option like -mprefer=agnostic to set this value???. */ - return MASK_ANY; + return TARGET_XTHEADVECTOR ? MASK_UNDISTURBED : MASK_ANY; } /* Get avl_type rtx. */ @@ -4383,7 +4383,7 @@ cmp_lmul_gt_one (machine_mode mode) bool vls_mode_valid_p (machine_mode vls_mode) { - if (!TARGET_VECTOR) + if (!TARGET_VECTOR || TARGET_XTHEADVECTOR) return false; if (riscv_autovec_preference == RVV_SCALABLE) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index c51affde353..22d877af9be 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -115,23 +115,37 @@ public: tree type = builtin_types[e.type.index].vector; machine_mode mode = TYPE_MODE (type); - /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. - - - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 - - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 - - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 - - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 - - e8,m2/e16,m4/e32,m8 --> e8m2 - - e8,m4/e16,m8 --> e8m4 - - e8,m8 --> e8m8 - */ - /* SEW. */ - e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); - - /* LMUL. */ - machine_mode e8_mode - = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); - e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + + if (TARGET_XTHEADVECTOR) + { + machine_mode inner_mode = GET_MODE_INNER (mode); + /* SEW. */ + e.add_input_operand (Pmode, + gen_int_mode (GET_MODE_BITSIZE (inner_mode), Pmode)); + /* LMUL. */ + e.add_input_operand (Pmode, + gen_int_mode (get_vlmul (mode), Pmode)); + } + else + { + /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. + + - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 + - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 + - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 + - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 + - e8,m2/e16,m4/e32,m8 --> e8m2 + - e8,m4/e16,m8 --> e8m4 + - e8,m8 --> e8m8 + */ + /* SEW. */ + e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); + + /* LMUL. */ + machine_mode e8_mode + = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); + e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + } /* TAIL_ANY. */ e.add_input_operand (Pmode, diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 4a754e0228f..6b49404a1fa 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -33,6 +33,25 @@ namespace riscv_vector { +/* Check whether the RETURN_TYPE and ARGUMENT_TYPES are + valid for the function. */ + +static bool +check_type (tree return_type, vec<tree> &argument_types) +{ + tree arg; + unsigned i; + + if (!return_type) + return false; + + FOR_EACH_VEC_ELT (argument_types, i, arg) + if (!arg) + return false; + + return true; +} + /* Add one function instance for GROUP, using operand suffix at index OI, mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ static void @@ -49,6 +68,10 @@ build_one (function_builder &b, const function_group_info &group, group.ops_infos.types[vec_type_idx].index); b.allocate_argument_types (function_instance, argument_types); b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + b.add_overloaded_function (function_instance, *group.shape); b.add_unique_function (function_instance, (*group.shape), return_type, argument_types); diff --git a/gcc/config/riscv/riscv-vector-switch.def b/gcc/config/riscv/riscv-vector-switch.def index 5c9f9bcbc3e..f7a66b34bae 100644 --- a/gcc/config/riscv/riscv-vector-switch.def +++ b/gcc/config/riscv/riscv-vector-switch.def @@ -68,9 +68,9 @@ Encode the ratio of SEW/LMUL into the mask types. #endif /* Disable modes if TARGET_MIN_VLEN == 32. */ -ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) -ENTRY (RVVMF32BI, true, LMUL_F4, 32) -ENTRY (RVVMF16BI, true, LMUL_F2, 16) +ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F8, 64) +ENTRY (RVVMF32BI, true, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F4, 32) +ENTRY (RVVMF16BI, true, TARGET_XTHEADVECTOR ? LMUL_1 : LMUL_F2 , 16) ENTRY (RVVMF8BI, true, LMUL_1, 8) ENTRY (RVVMF4BI, true, LMUL_2, 4) ENTRY (RVVMF2BI, true, LMUL_4, 2) @@ -81,39 +81,39 @@ ENTRY (RVVM8QI, true, LMUL_8, 1) ENTRY (RVVM4QI, true, LMUL_4, 2) ENTRY (RVVM2QI, true, LMUL_2, 4) ENTRY (RVVM1QI, true, LMUL_1, 8) -ENTRY (RVVMF2QI, true, LMUL_F2, 16) -ENTRY (RVVMF4QI, true, LMUL_F4, 32) -ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) +ENTRY (RVVMF2QI, !TARGET_XTHEADVECTOR, LMUL_F2, 16) +ENTRY (RVVMF4QI, !TARGET_XTHEADVECTOR, LMUL_F4, 32) +ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F8, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8HI, true, LMUL_8, 2) ENTRY (RVVM4HI, true, LMUL_4, 4) ENTRY (RVVM2HI, true, LMUL_2, 8) ENTRY (RVVM1HI, true, LMUL_1, 16) -ENTRY (RVVMF2HI, true, LMUL_F2, 32) -ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HI, !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_16. */ ENTRY (RVVM8HF, TARGET_VECTOR_ELEN_FP_16, LMUL_8, 2) ENTRY (RVVM4HF, TARGET_VECTOR_ELEN_FP_16, LMUL_4, 4) ENTRY (RVVM2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_2, 8) ENTRY (RVVM1HF, TARGET_VECTOR_ELEN_FP_16, LMUL_1, 16) -ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_F2, 32) -ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8SI, true, LMUL_8, 4) ENTRY (RVVM4SI, true, LMUL_4, 8) ENTRY (RVVM2SI, true, LMUL_2, 16) ENTRY (RVVM1SI, true, LMUL_1, 32) -ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_32. */ ENTRY (RVVM8SF, TARGET_VECTOR_ELEN_FP_32, LMUL_8, 4) ENTRY (RVVM4SF, TARGET_VECTOR_ELEN_FP_32, LMUL_4, 8) ENTRY (RVVM2SF, TARGET_VECTOR_ELEN_FP_32, LMUL_2, 16) ENTRY (RVVM1SF, TARGET_VECTOR_ELEN_FP_32, LMUL_1, 32) -ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if !TARGET_VECTOR_ELEN_64. */ ENTRY (RVVM8DI, TARGET_VECTOR_ELEN_64, LMUL_8, 8) @@ -140,127 +140,127 @@ ENTRY (RVVM1DF, TARGET_VECTOR_ELEN_FP_64, LMUL_1, 64) #endif TUPLE_ENTRY (RVVM1x8QI, true, RVVM1QI, 8, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x8QI, true, RVVMF2QI, 8, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x8QI, true, RVVMF4QI, 8, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 8, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x8QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 8, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x8QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 8, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 8, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x7QI, true, RVVM1QI, 7, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x7QI, true, RVVMF2QI, 7, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x7QI, true, RVVMF4QI, 7, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 7, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x7QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 7, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x7QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 7, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 7, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x6QI, true, RVVM1QI, 6, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x6QI, true, RVVMF2QI, 6, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x6QI, true, RVVMF4QI, 6, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 6, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x6QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 6, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x6QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 6, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 6, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x5QI, true, RVVM1QI, 5, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x5QI, true, RVVMF2QI, 5, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x5QI, true, RVVMF4QI, 5, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 5, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x5QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 5, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x5QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 5, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 5, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x4QI, true, RVVM2QI, 4, LMUL_2, 4) TUPLE_ENTRY (RVVM1x4QI, true, RVVM1QI, 4, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x4QI, true, RVVMF2QI, 4, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x4QI, true, RVVMF4QI, 4, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 4, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x4QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 4, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x4QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 4, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 4, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x3QI, true, RVVM2QI, 3, LMUL_2, 4) TUPLE_ENTRY (RVVM1x3QI, true, RVVM1QI, 3, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x3QI, true, RVVMF2QI, 3, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x3QI, true, RVVMF4QI, 3, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 3, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x3QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 3, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x3QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 3, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 3, LMUL_F8, 64) TUPLE_ENTRY (RVVM4x2QI, true, RVVM4QI, 2, LMUL_4, 2) TUPLE_ENTRY (RVVM2x2QI, true, RVVM2QI, 2, LMUL_2, 4) TUPLE_ENTRY (RVVM1x2QI, true, RVVM1QI, 2, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x2QI, true, RVVMF2QI, 2, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x2QI, true, RVVMF4QI, 2, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 2, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x2QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 2, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x2QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 2, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 2, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x8HI, true, RVVM1HI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HI, true, RVVMF2HI, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HI, true, RVVM1HI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HI, true, RVVMF2HI, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HI, true, RVVM1HI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HI, true, RVVMF2HI, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HI, true, RVVM1HI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HI, true, RVVMF2HI, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HI, true, RVVM2HI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HI, true, RVVM1HI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HI, true, RVVMF2HI, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HI, true, RVVM2HI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HI, true, RVVM1HI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HI, true, RVVMF2HI, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HI, true, RVVM4HI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HI, true, RVVM2HI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HI, true, RVVM1HI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HI, true, RVVMF2HI, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM4HF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8SI, true, RVVM1SI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SI, true, RVVM1SI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SI, true, RVVM1SI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SI, true, RVVM1SI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SI, true, RVVM2SI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SI, true, RVVM1SI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SI, true, RVVM2SI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SI, true, RVVM1SI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SI, true, RVVM4SI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SI, true, RVVM2SI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SI, true, RVVM1SI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM4SF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 8, LMUL_1, 16) TUPLE_ENTRY (RVVM1x7DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 7, LMUL_1, 16) diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index 51878797287..44a36db7803 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -1406,6 +1406,9 @@ riscv_v_adjust_bytesize (machine_mode mode, int scale) { if (riscv_v_ext_vector_mode_p (mode)) { + if (TARGET_XTHEADVECTOR) + return BYTES_PER_RISCV_VECTOR; + poly_int64 nunits = GET_MODE_NUNITS (mode); poly_int64 mode_size = GET_MODE_SIZE (mode); @@ -9962,7 +9965,7 @@ riscv_use_divmod_expander (void) static machine_mode riscv_preferred_simd_mode (scalar_mode mode) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::preferred_simd_mode (mode); return word_mode; @@ -10313,7 +10316,7 @@ riscv_mode_priority (int, int n) unsigned int riscv_autovectorize_vector_modes (vector_modes *modes, bool all) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::autovectorize_vector_modes (modes, all); return default_autovectorize_vector_modes (modes, all); @@ -10496,6 +10499,16 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) return false; } +/* Implements target hook vector_mode_supported_any_target_p. */ + +static bool +riscv_vector_mode_supported_any_target_p (machine_mode mode) +{ + if (TARGET_XTHEADVECTOR) + return false; + return true; +} + /* Initialize the GCC target structure. */ #undef TARGET_ASM_ALIGNED_HI_OP #define TARGET_ASM_ALIGNED_HI_OP "\t.half\t" @@ -10839,6 +10852,9 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) #undef TARGET_PREFERRED_ELSE_VALUE #define TARGET_PREFERRED_ELSE_VALUE riscv_preferred_else_value +#undef TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P +#define TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P riscv_vector_mode_supported_any_target_p + struct gcc_target targetm = TARGET_INITIALIZER; #include "gt-riscv.h" diff --git a/gcc/config/riscv/riscv_th_vector.h b/gcc/config/riscv/riscv_th_vector.h new file mode 100644 index 00000000000..a897b9cb90a --- /dev/null +++ b/gcc/config/riscv/riscv_th_vector.h @@ -0,0 +1,49 @@ +/* RISC-V 'XTheadVector' Extension intrinsics include file. + Copyright (C) 2022-2023 Free Software Foundation, Inc. + + This file is part of GCC. + + GCC is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published + by the Free Software Foundation; either version 3, or (at your + option) any later version. + + GCC is distributed in the hope that it will be useful, but WITHOUT + ANY WARRANTY; without even the implied warranty of MERCHANTABILITY + or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public + License for more details. + + Under Section 7 of GPL version 3, you are granted additional + permissions described in the GCC Runtime Library Exception, version + 3.1, as published by the Free Software Foundation. + + You should have received a copy of the GNU General Public License and + a copy of the GCC Runtime Library Exception along with this program; + see the files COPYING3 and COPYING.RUNTIME respectively. If not, see + <" target="_blank">http://www.gnu.org/licenses/>. */ + +#ifndef __RISCV_TH_VECTOR_H +#define __RISCV_TH_VECTOR_H + +#include <stdint.h> +#include <stddef.h> + +#ifndef __riscv_xtheadvector +#error "XTheadVector intrinsics require the xtheadvector extension." +#else +#ifdef __cplusplus +extern "C" { +#endif + +/* NOTE: This implementation of riscv_th_vector.h is intentionally short. It does + not define the RVV types and intrinsic functions directly in C and C++ + code, but instead uses the following pragma to tell GCC to insert the + necessary type and function definitions itself. The net effect is the + same, and the file is a complete implementation of riscv_th_vector.h. */ +#pragma riscv intrinsic "xtheadvector" + +#ifdef __cplusplus +} +#endif // __cplusplus +#endif // __riscv_xtheadvector +#endif // __RISCV_TH_ECTOR_H diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md new file mode 100644 index 00000000000..1230bfb5994 --- /dev/null +++ b/gcc/config/riscv/thead-vector.md @@ -0,0 +1,69 @@ +(define_c_enum "unspec" [ + UNSPEC_TH_VWLDST +]) + +(define_mode_iterator V_VLS_VT [V VLS VT]) +(define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) + +(define_split + [(set (match_operand:V_VB_VLS_VT 0 "reg_or_mem_operand") + (match_operand:V_VB_VLS_VT 1 "reg_or_mem_operand"))] + "TARGET_XTHEADVECTOR" + [(const_int 0)] + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + }) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:V_VLS_VT 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:V_VLS_VT + [(match_operand:V_VLS_VT 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2")]) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:VB 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:VB + [(match_operand:VB 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2") + (set (attr "sew") (const_int 8)) + (set (attr "vlmul") (symbol_ref "riscv_vector::LMUL_1"))]) diff --git a/gcc/config/riscv/vector-iterators.md b/gcc/config/riscv/vector-iterators.md index 5f5f7b5b986..c0fc7a2441d 100644 --- a/gcc/config/riscv/vector-iterators.md +++ b/gcc/config/riscv/vector-iterators.md @@ -109,11 +109,11 @@ ]) (define_mode_iterator VI [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -128,11 +128,11 @@ ;; allow the instruction and mode to be matched during combine et al. (define_mode_iterator VF [ (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -140,11 +140,11 @@ (define_mode_iterator VF_ZVFHMIN [ (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -271,16 +271,16 @@ ]) (define_mode_iterator VEEWEXT2 [ - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -290,10 +290,10 @@ ]) (define_mode_iterator VEEWEXT4 [ - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -311,59 +311,59 @@ ]) (define_mode_iterator VEEWTRUNC2 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") (RVVM4SI "TARGET_64BIT") (RVVM2SI "TARGET_64BIT") (RVVM1SI "TARGET_64BIT") - (RVVMF2SI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEEWTRUNC4 [ - RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM2HI "TARGET_64BIT") (RVVM1HI "TARGET_64BIT") - (RVVMF2HI "TARGET_64BIT") - (RVVMF4HI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2HI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") (RVVM1HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEEWTRUNC8 [ (RVVM1QI "TARGET_64BIT") - (RVVMF2QI "TARGET_64BIT") - (RVVMF4QI "TARGET_64BIT") - (RVVMF8QI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2QI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF4QI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEI16 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -452,11 +452,11 @@ ]) (define_mode_iterator VFULLI [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_FULL_V") (RVVM4DI "TARGET_FULL_V") (RVVM2DI "TARGET_FULL_V") (RVVM1DI "TARGET_FULL_V") @@ -509,17 +509,17 @@ ]) (define_mode_iterator VI_QH [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") ]) (define_mode_iterator VI_QHS [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -560,11 +560,11 @@ ]) (define_mode_iterator VI_QHS_NO_M8 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -603,11 +603,11 @@ (define_mode_iterator VF_HS [ (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (V1HF "riscv_vector::vls_mode_valid_p (V1HFmode) && TARGET_ZVFH") (V2HF "riscv_vector::vls_mode_valid_p (V2HFmode) && TARGET_ZVFH") @@ -638,12 +638,12 @@ (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") - (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (V1HF "riscv_vector::vls_mode_valid_p (V1HFmode) && TARGET_ZVFH") (V2HF "riscv_vector::vls_mode_valid_p (V2HFmode) && TARGET_ZVFH") @@ -674,11 +674,11 @@ ]) (define_mode_iterator V_VLSI_QHS [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -756,27 +756,27 @@ ;; E.g. when index mode = RVVM8QImde and Pmode = SImode, if it is not zero_extend or ;; scalar != 1, such gather/scatter is not allowed since we don't have RVVM32SImode. (define_mode_iterator RATIO64 [ - (RVVMF8QI "TARGET_MIN_VLEN > 32") - (RVVMF4HI "TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM1DI "TARGET_VECTOR_ELEN_64") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") ]) (define_mode_iterator RATIO32 [ - RVVMF4QI - RVVMF2HI + (RVVMF4QI "!TARGET_XTHEADVECTOR") + (RVVMF2HI "!TARGET_XTHEADVECTOR") RVVM1SI (RVVM2DI "TARGET_VECTOR_ELEN_64") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") ]) (define_mode_iterator RATIO16 [ - RVVMF2QI + (RVVMF2QI "!TARGET_XTHEADVECTOR") RVVM1HI RVVM2SI (RVVM4DI "TARGET_VECTOR_ELEN_64") @@ -814,21 +814,21 @@ ]) (define_mode_iterator RATIO64I [ - (RVVMF8QI "TARGET_MIN_VLEN > 32") - (RVVMF4HI "TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM1DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") ]) (define_mode_iterator RATIO32I [ - RVVMF4QI - RVVMF2HI + (RVVMF4QI "!TARGET_XTHEADVECTOR") + (RVVMF2HI "!TARGET_XTHEADVECTOR") RVVM1SI (RVVM2DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") ]) (define_mode_iterator RATIO16I [ - RVVMF2QI + (RVVMF2QI "!TARGET_XTHEADVECTOR") RVVM1HI RVVM2SI (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") @@ -873,21 +873,21 @@ ]) (define_mode_iterator V_FRACT [ - RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") ]) (define_mode_iterator VWEXTI [ - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -933,7 +933,7 @@ (RVVM4SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -966,7 +966,7 @@ (RVVM4SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -996,7 +996,7 @@ (define_mode_iterator VWCONVERTI [ (RVVM8SI "TARGET_ZVFH") (RVVM4SI "TARGET_ZVFH") (RVVM2SI "TARGET_ZVFH") (RVVM1SI "TARGET_ZVFH") - (RVVMF2SI "TARGET_ZVFH") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_ZVFH") (RVVM8DI "TARGET_VECTOR_ELEN_64 && TARGET_VECTOR_ELEN_FP_32") (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_VECTOR_ELEN_FP_32") @@ -1045,7 +1045,7 @@ ]) (define_mode_iterator VQEXTI [ - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -1456,11 +1456,11 @@ ;; VINDEXED [VI8 VI16 VI32 (VI64 "TARGET_64BIT")]. (define_mode_iterator VINDEXED [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") @@ -1468,12 +1468,12 @@ (RVVM1DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64 && TARGET_64BIT") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64 && TARGET_64BIT") @@ -3173,11 +3173,11 @@ (define_mode_iterator V_VLS_F_CONVERT_SI [ (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") - (RVVMF2HF "TARGET_ZVFH") (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") @@ -3290,12 +3290,12 @@ ]) (define_mode_iterator V_VLS_F_CONVERT_DI [ - (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index b5a9055cdc4..cb30c9ae97c 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -83,7 +83,7 @@ ;; check. However, we need default value of SEW for vsetvl instruction since there ;; is no field for ratio in the vsetvl instruction encoding. (define_attr "sew" "" - (cond [(eq_attr "mode" "RVVMF64BI,RVVMF32BI,RVVMF16BI,RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ + (cond [(eq_attr "mode" "RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ RVVM8QI,RVVM4QI,RVVM2QI,RVVM1QI,RVVMF2QI,RVVMF4QI,RVVMF8QI,\ RVVM1x8QI,RVVMF2x8QI,RVVMF4x8QI,RVVMF8x8QI,\ RVVM1x7QI,RVVMF2x7QI,RVVMF4x7QI,RVVMF8x7QI,\ @@ -95,6 +95,18 @@ V1QI,V2QI,V4QI,V8QI,V16QI,V32QI,V64QI,V128QI,V256QI,V512QI,V1024QI,V2048QI,V4096QI,\ V1BI,V2BI,V4BI,V8BI,V16BI,V32BI,V64BI,V128BI,V256BI,V512BI,V1024BI,V2048BI,V4096BI") (const_int 8) + (eq_attr "mode" "RVVMF16BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 16) + (const_int 8)) + (eq_attr "mode" "RVVMF32BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 32) + (const_int 8)) + (eq_attr "mode" "RVVMF64BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 64) + (const_int 8)) (eq_attr "mode" "RVVM8HI,RVVM4HI,RVVM2HI,RVVM1HI,RVVMF2HI,RVVMF4HI,\ RVVM1x8HI,RVVMF2x8HI,RVVMF4x8HI,\ RVVM1x7HI,RVVMF2x7HI,RVVMF4x7HI,\ @@ -155,9 +167,9 @@ (eq_attr "mode" "RVVM4QI,RVVMF2BI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2QI,RVVMF4BI") (symbol_ref "riscv_vector::LMUL_2") (eq_attr "mode" "RVVM1QI,RVVMF8BI") (symbol_ref "riscv_vector::LMUL_1") - (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "riscv_vector::LMUL_F2") - (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "riscv_vector::LMUL_F4") - (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "riscv_vector::LMUL_F8") + (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F2") + (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F4") + (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F8") (eq_attr "mode" "RVVM8HI") (symbol_ref "riscv_vector::LMUL_8") (eq_attr "mode" "RVVM4HI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2HI") (symbol_ref "riscv_vector::LMUL_2") @@ -428,6 +440,10 @@ vislide1up,vislide1down,vfslide1up,vfslide1down,\ vgather,vcompress,vlsegdux,vlsegdox,vssegtux,vssegtox") (const_int INVALID_ATTRIBUTE) + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE) (eq_attr "mode" "RVVM8QI,RVVM1BI") (const_int 1) (eq_attr "mode" "RVVM4QI,RVVMF2BI") (const_int 2) (eq_attr "mode" "RVVM2QI,RVVMF4BI") (const_int 4) @@ -888,6 +904,8 @@ (symbol_ref "riscv_vector::FRM_DYN")] (symbol_ref "riscv_vector::FRM_NONE"))) +(include "thead-vector.md") + ;; ----------------------------------------------------------------- ;; ---- Miscellaneous Operations ;; ----------------------------------------------------------------- @@ -1061,6 +1079,12 @@ - We can not leave it to TARGET_SECONDARY_RELOAD since it happens before spilling. The clobber scratch is used by spilling fractional registers in IRA/LRA so it's too early. */ + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; @@ -1097,7 +1121,7 @@ (define_insn "*mov<mode>_whole" [(set (match_operand:V_WHOLE 0 "reg_or_mem_operand" "=vr, m,vr") (match_operand:V_WHOLE 1 "reg_or_mem_operand" " m,vr,vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "@ vl%m1re<sew>.v\t%0,%1 vs%m1r.v\t%1,%0 @@ -1118,6 +1142,13 @@ (match_operand:VB 1 "general_operand"))] "TARGET_VECTOR" { + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } + if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; }) @@ -1125,7 +1156,7 @@ (define_insn "*mov<mode>" [(set (match_operand:VB 0 "register_operand" "=vr") (match_operand:VB 1 "register_operand" " vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "vmv1r.v\t%0,%1" [(set_attr "type" "vmov") (set_attr "mode" "<MODE>")]) @@ -1511,7 +1542,7 @@ (match_dup 4) (match_dup 5)] UNSPEC_VSETVL))] "TARGET_VECTOR" - "vset%i1vli\t%0,%1,e%2,%m3,t%p4,m%p5" + { return TARGET_XTHEADVECTOR ? "vsetvli\t%0,%1,e%2,%m3" : "vset%i1vli\t%0,%1,e%2,%m3,t%p4,m%p5"; } [(set_attr "type" "vsetvl") (set_attr "mode" "<MODE>") (set (attr "sew") (symbol_ref "INTVAL (operands[2])")) @@ -1529,7 +1560,7 @@ (match_operand 2 "const_int_operand" "i") (match_operand 3 "const_int_operand" "i")] UNSPEC_VSETVL))] "TARGET_VECTOR" - "vsetvli\tzero,zero,e%0,%m1,t%p2,m%p3" + { return TARGET_XTHEADVECTOR ? "vsetvli\tzero,zero,e%0,%m1" : "vsetvli\tzero,zero,e%0,%m1,t%p2,m%p3"; } [(set_attr "type" "vsetvl") (set_attr "mode" "SI") (set (attr "sew") (symbol_ref "INTVAL (operands[0])")) @@ -1551,7 +1582,7 @@ (match_operand 3 "const_int_operand" "i") (match_operand 4 "const_int_operand" "i")] UNSPEC_VSETVL))] "TARGET_VECTOR" - "vset%i0vli\tzero,%0,e%1,%m2,t%p3,m%p4" + { return TARGET_XTHEADVECTOR ? "vsetvli\tzero,%0,e%1,%m2" : "vset%i0vli\tzero,%0,e%1,%m2,t%p3,m%p4"; } [(set_attr "type" "vsetvl") (set_attr "mode" "<MODE>") (set (attr "sew") (symbol_ref "INTVAL (operands[1])")) @@ -3680,7 +3711,7 @@ (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3701,7 +3732,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3722,7 +3753,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c index 2e0e12aa045..2eef9e1e1a8 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c @@ -1,4 +1,4 @@ -/* { dg-do compile } */ +/* { dg-do compile { target { ! riscv_xtheadvector } } } */ /* { dg-skip-if "test rvv intrinsic" { *-*-* } { "*" } { "-march=rv*v*" } } */ void foo0 () {__rvv_bool64_t t;} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c index 3d81b179235..ef329e30785 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c @@ -1,4 +1,4 @@ /* { dg-do compile } */ /* { dg-options "-O3 -march=rv32gc -mabi=ilp32d" } */ -#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' extension enabled} } */ +#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' or 'XTHEADVECTOR' extension enabled} } */ diff --git a/gcc/testsuite/lib/target-supports.exp b/gcc/testsuite/lib/target-supports.exp index 7f13ff0ca56..70df6b1401c 100644 --- a/gcc/testsuite/lib/target-supports.exp +++ b/gcc/testsuite/lib/target-supports.exp @@ -1952,6 +1952,18 @@ proc check_effective_target_riscv_zbb { } { }] } +# Return 1 if the target arch supports the XTheadVector extension, 0 otherwise. +# Cache the result. + +proc check_effective_target_riscv_xtheadvector { } { + return [check_no_compiler_messages riscv_ext_xtheadvector assembly { + #ifndef __riscv_xtheadvector + #error "Not __riscv_xtheadvector" + #endif + }] +} + + # Return 1 if we can execute code when using dg-add-options riscv_v proc check_effective_target_riscv_v_ok { } { -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re：[PATCH v4] RISC-V: Handle differences between XTheadvector and Vector 2024-01-08 23:08 ` [PATCH v4] " 钟居哲 2024-01-09 2:12 ` Re：[PATCH " joshua @ 2024-01-09 3:23 ` joshua 1 sibling, 0 replies; 133+ messages in thread From: joshua @ 2024-01-09 3:23 UTC (permalink / raw) To: 钟居哲, gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, Jeff Law, Christoph Müllner, jinma, Cooper Qu It has been updated. [PATCH v5] RISC-V: Handle differences between XTheadvector and Vector (gnu.org) ------------------------------------------------------------------ 发件人：钟居哲 <juzhe.zhong@rivai.ai> 发送时间：2024年1月9日(星期二) 07:08 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送："jim.wilson.gcc"<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; Jeff Law<jeffreyalaw@gmail.com>; "Christoph Müllner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; Cooper Qu<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v4] RISC-V: Handle differences between XTheadvector and Vector - return TAIL_ANY; + return TARGET_XTHEADVECTOR ? TAIL_AGNOSTIC : TAIL_ANY; - return MASK_ANY; + return TARGET_XTHEADVECTOR ? MASK_UNDISTURBED : MASK_ANY; You shouldn't change this. - "vset%i1vli\t%0,%1,e%2,%m3,t%p4,m%p5" + { return TARGET_XTHEADVECTOR ? "vsetvli\t%0,%1,e%2,%m3" : "vset%i1vli\t%0,%1,e%2,%m3,t%p4,m%p5"; } I prefer do it in ASM_OUTPUT + Copyright (C) 2022-2023 Free Software Foundation, Inc. Copyright is not correct. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-03 14:15 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v4] RISC-V: Handle differences between XTheadvector and Vector This patch is to handle the differences in instruction generation between Vector and XTheadVector. In this version, we only support partial xtheadvector instructions that leverage directly from current RVV1.0 with simple adding "th." prefix. For different name xtheadvector instructions but share same patterns as RVV1.0 instructions, we will use ASM targethook to rewrite the whole string of the instructions in the following patches. For some vector patterns that cannot be avoided, we use "!TARGET_XTHEADVECTOR" to disable them in vector.md in order not to generate instructions that xtheadvector does not support, like vmv1r and vsext.vf2. gcc/ChangeLog: * config.gcc: Add files for XTheadVector intrinsics. * config/riscv/autovec.md: Guard XTheadVector. * config/riscv/riscv-c.cc: Add pragma for XTheadVector. * config/riscv/riscv-string.cc (expand_block_move): Guard XTheadVector. (get_prefer_tail_policy): Give specific value for tail. (get_prefer_mask_policy): Give specific value for mask. (vls_mode_valid_p): Avoid autovec. * config/riscv/riscv-vector-builtins-shapes.cc (check_type): (build_one): New function. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): (DEF_THEAD_RVV_FUNCTION): Add new marcos. (check_required_extensions): (handle_pragma_vector): * config/riscv/riscv-vector-builtins.h (RVV_REQUIRE_VECTOR): (RVV_REQUIRE_XTHEADVECTOR): Add RVV_REQUIRE_VECTOR and RVV_REQUIRE_XTHEADVECTOR. (struct function_group_info): * config/riscv/riscv-vector-switch.def (ENTRY): Disable fractional mode for the XTheadVector extension. (TUPLE_ENTRY): Likewise. * config/riscv/riscv-vsetvl.cc: Add functions for xtheadvector. * config/riscv/riscv.cc (riscv_v_ext_vls_mode_p): Guard XTheadVector. (riscv_v_adjust_bytesize): Likewise. (riscv_preferred_simd_mode): Likewsie. (riscv_autovectorize_vector_modes): Likewise. (riscv_vector_mode_supported_any_target_p): Likewise. (TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P): Likewise. * config/riscv/vector-iterators.md: Remove fractional LMUL. * config/riscv/vector.md: Include thead-vector.md. * config/riscv/riscv_th_vector.h: New file. * config/riscv/thead-vector.md: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/base/pragma-1.c: Add XTheadVector. * gcc.target/riscv/rvv/base/abi-1.c: Exclude XTheadVector. * lib/target-supports.exp: Add target for XTheadVector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config.gcc | 2 +- gcc/config/riscv/autovec.md | 2 +- gcc/config/riscv/predicates.md | 4 +- gcc/config/riscv/riscv-c.cc | 3 +- gcc/config/riscv/riscv-string.cc | 3 +- gcc/config/riscv/riscv-v.cc | 6 +- .../riscv/riscv-vector-builtins-bases.cc | 48 +++-- .../riscv/riscv-vector-builtins-shapes.cc | 23 +++ gcc/config/riscv/riscv-vector-switch.def | 150 +++++++------- gcc/config/riscv/riscv.cc | 20 +- gcc/config/riscv/riscv_th_vector.h | 49 +++++ gcc/config/riscv/thead-vector.md | 69 +++++++ gcc/config/riscv/vector-iterators.md | 186 +++++++++--------- gcc/config/riscv/vector.md | 55 ++++-- .../gcc.target/riscv/rvv/base/abi-1.c | 2 +- .../gcc.target/riscv/rvv/base/pragma-1.c | 2 +- gcc/testsuite/lib/target-supports.exp | 12 ++ 17 files changed, 427 insertions(+), 209 deletions(-) create mode 100644 gcc/config/riscv/riscv_th_vector.h create mode 100644 gcc/config/riscv/thead-vector.md diff --git a/gcc/config.gcc b/gcc/config.gcc index f0676c830e8..1445d98c147 100644 --- a/gcc/config.gcc +++ b/gcc/config.gcc @@ -549,7 +549,7 @@ riscv*) extra_objs="${extra_objs} riscv-vector-builtins.o riscv-vector-builtins-shapes.o riscv-vector-builtins-bases.o" extra_objs="${extra_objs} thead.o riscv-target-attr.o" d_target_objs="riscv-d.o" - extra_headers="riscv_vector.h" + extra_headers="riscv_vector.h riscv_th_vector.h" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.cc" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.h" ;; diff --git a/gcc/config/riscv/autovec.md b/gcc/config/riscv/autovec.md index 8b8a92f10a1..1fac56c7095 100644 --- a/gcc/config/riscv/autovec.md +++ b/gcc/config/riscv/autovec.md @@ -2579,7 +2579,7 @@ [(match_operand 0 "register_operand") (match_operand 1 "memory_operand") (match_operand:ANYI 2 "const_int_operand")] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" { riscv_vector::expand_rawmemchr(<MODE>mode, operands[0], operands[1], operands[2]); diff --git a/gcc/config/riscv/predicates.md b/gcc/config/riscv/predicates.md index b87a6900841..6e36c5afae1 100644 --- a/gcc/config/riscv/predicates.md +++ b/gcc/config/riscv/predicates.md @@ -428,7 +428,9 @@ ;; Predicates for the V extension. (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (ior (match_test "TARGET_XTHEADVECTOR && rtx_equal_p (op, const0_rtx)") + (match_test "!TARGET_XTHEADVECTOR")) + (match_operand 0 "const_csr_operand")))) (define_special_predicate "autovec_length_operand" (ior (match_operand 0 "pmode_register_operand") diff --git a/gcc/config/riscv/riscv-c.cc b/gcc/config/riscv/riscv-c.cc index d290c58613a..d6fb83ac00c 100644 --- a/gcc/config/riscv/riscv-c.cc +++ b/gcc/config/riscv/riscv-c.cc @@ -195,7 +195,8 @@ riscv_pragma_intrinsic (cpp_reader *) const char *name = TREE_STRING_POINTER (x); - if (strcmp (name, "vector") == 0) + if (strcmp (name, "vector") == 0 + || strcmp (name, "xtheadvector") == 0) { if (!TARGET_VECTOR) { diff --git a/gcc/config/riscv/riscv-string.cc b/gcc/config/riscv/riscv-string.cc index 11c1f74d0b3..f431b562222 100644 --- a/gcc/config/riscv/riscv-string.cc +++ b/gcc/config/riscv/riscv-string.cc @@ -773,7 +773,8 @@ riscv_expand_block_move_scalar (rtx dest, rtx src, rtx length) bool riscv_expand_block_move (rtx dest, rtx src, rtx length) { - if (TARGET_VECTOR && stringop_strategy & STRATEGY_VECTOR) + if ((TARGET_VECTOR && !TARGET_XTHEADVECTOR) + && stringop_strategy & STRATEGY_VECTOR) { bool ok = riscv_vector::expand_block_move (dest, src, length); if (ok) diff --git a/gcc/config/riscv/riscv-v.cc b/gcc/config/riscv/riscv-v.cc index 038ab084a37..eb061b593a5 100644 --- a/gcc/config/riscv/riscv-v.cc +++ b/gcc/config/riscv/riscv-v.cc @@ -1772,7 +1772,7 @@ get_prefer_tail_policy () compiler pick up either agnostic or undisturbed. Maybe we will have a compile option like -mprefer=agnostic to set this value???. */ - return TAIL_ANY; + return TARGET_XTHEADVECTOR ? TAIL_AGNOSTIC : TAIL_ANY; } /* Get prefer mask policy. */ @@ -1783,7 +1783,7 @@ get_prefer_mask_policy () compiler pick up either agnostic or undisturbed. Maybe we will have a compile option like -mprefer=agnostic to set this value???. */ - return MASK_ANY; + return TARGET_XTHEADVECTOR ? MASK_UNDISTURBED : MASK_ANY; } /* Get avl_type rtx. */ @@ -4383,7 +4383,7 @@ cmp_lmul_gt_one (machine_mode mode) bool vls_mode_valid_p (machine_mode vls_mode) { - if (!TARGET_VECTOR) + if (!TARGET_VECTOR || TARGET_XTHEADVECTOR) return false; if (riscv_autovec_preference == RVV_SCALABLE) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index c51affde353..22d877af9be 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -115,23 +115,37 @@ public: tree type = builtin_types[e.type.index].vector; machine_mode mode = TYPE_MODE (type); - /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. - - - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 - - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 - - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 - - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 - - e8,m2/e16,m4/e32,m8 --> e8m2 - - e8,m4/e16,m8 --> e8m4 - - e8,m8 --> e8m8 - */ - /* SEW. */ - e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); - - /* LMUL. */ - machine_mode e8_mode - = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); - e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + + if (TARGET_XTHEADVECTOR) + { + machine_mode inner_mode = GET_MODE_INNER (mode); + /* SEW. */ + e.add_input_operand (Pmode, + gen_int_mode (GET_MODE_BITSIZE (inner_mode), Pmode)); + /* LMUL. */ + e.add_input_operand (Pmode, + gen_int_mode (get_vlmul (mode), Pmode)); + } + else + { + /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. + + - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 + - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 + - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 + - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 + - e8,m2/e16,m4/e32,m8 --> e8m2 + - e8,m4/e16,m8 --> e8m4 + - e8,m8 --> e8m8 + */ + /* SEW. */ + e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); + + /* LMUL. */ + machine_mode e8_mode + = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); + e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + } /* TAIL_ANY. */ e.add_input_operand (Pmode, diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 4a754e0228f..6b49404a1fa 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -33,6 +33,25 @@ namespace riscv_vector { +/* Check whether the RETURN_TYPE and ARGUMENT_TYPES are + valid for the function. */ + +static bool +check_type (tree return_type, vec<tree> &argument_types) +{ + tree arg; + unsigned i; + + if (!return_type) + return false; + + FOR_EACH_VEC_ELT (argument_types, i, arg) + if (!arg) + return false; + + return true; +} + /* Add one function instance for GROUP, using operand suffix at index OI, mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ static void @@ -49,6 +68,10 @@ build_one (function_builder &b, const function_group_info &group, group.ops_infos.types[vec_type_idx].index); b.allocate_argument_types (function_instance, argument_types); b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + b.add_overloaded_function (function_instance, *group.shape); b.add_unique_function (function_instance, (*group.shape), return_type, argument_types); diff --git a/gcc/config/riscv/riscv-vector-switch.def b/gcc/config/riscv/riscv-vector-switch.def index 5c9f9bcbc3e..f7a66b34bae 100644 --- a/gcc/config/riscv/riscv-vector-switch.def +++ b/gcc/config/riscv/riscv-vector-switch.def @@ -68,9 +68,9 @@ Encode the ratio of SEW/LMUL into the mask types. #endif /* Disable modes if TARGET_MIN_VLEN == 32. */ -ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) -ENTRY (RVVMF32BI, true, LMUL_F4, 32) -ENTRY (RVVMF16BI, true, LMUL_F2, 16) +ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F8, 64) +ENTRY (RVVMF32BI, true, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F4, 32) +ENTRY (RVVMF16BI, true, TARGET_XTHEADVECTOR ? LMUL_1 : LMUL_F2 , 16) ENTRY (RVVMF8BI, true, LMUL_1, 8) ENTRY (RVVMF4BI, true, LMUL_2, 4) ENTRY (RVVMF2BI, true, LMUL_4, 2) @@ -81,39 +81,39 @@ ENTRY (RVVM8QI, true, LMUL_8, 1) ENTRY (RVVM4QI, true, LMUL_4, 2) ENTRY (RVVM2QI, true, LMUL_2, 4) ENTRY (RVVM1QI, true, LMUL_1, 8) -ENTRY (RVVMF2QI, true, LMUL_F2, 16) -ENTRY (RVVMF4QI, true, LMUL_F4, 32) -ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) +ENTRY (RVVMF2QI, !TARGET_XTHEADVECTOR, LMUL_F2, 16) +ENTRY (RVVMF4QI, !TARGET_XTHEADVECTOR, LMUL_F4, 32) +ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F8, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8HI, true, LMUL_8, 2) ENTRY (RVVM4HI, true, LMUL_4, 4) ENTRY (RVVM2HI, true, LMUL_2, 8) ENTRY (RVVM1HI, true, LMUL_1, 16) -ENTRY (RVVMF2HI, true, LMUL_F2, 32) -ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HI, !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_16. */ ENTRY (RVVM8HF, TARGET_VECTOR_ELEN_FP_16, LMUL_8, 2) ENTRY (RVVM4HF, TARGET_VECTOR_ELEN_FP_16, LMUL_4, 4) ENTRY (RVVM2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_2, 8) ENTRY (RVVM1HF, TARGET_VECTOR_ELEN_FP_16, LMUL_1, 16) -ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_F2, 32) -ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8SI, true, LMUL_8, 4) ENTRY (RVVM4SI, true, LMUL_4, 8) ENTRY (RVVM2SI, true, LMUL_2, 16) ENTRY (RVVM1SI, true, LMUL_1, 32) -ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_32. */ ENTRY (RVVM8SF, TARGET_VECTOR_ELEN_FP_32, LMUL_8, 4) ENTRY (RVVM4SF, TARGET_VECTOR_ELEN_FP_32, LMUL_4, 8) ENTRY (RVVM2SF, TARGET_VECTOR_ELEN_FP_32, LMUL_2, 16) ENTRY (RVVM1SF, TARGET_VECTOR_ELEN_FP_32, LMUL_1, 32) -ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if !TARGET_VECTOR_ELEN_64. */ ENTRY (RVVM8DI, TARGET_VECTOR_ELEN_64, LMUL_8, 8) @@ -140,127 +140,127 @@ ENTRY (RVVM1DF, TARGET_VECTOR_ELEN_FP_64, LMUL_1, 64) #endif TUPLE_ENTRY (RVVM1x8QI, true, RVVM1QI, 8, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x8QI, true, RVVMF2QI, 8, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x8QI, true, RVVMF4QI, 8, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 8, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x8QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 8, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x8QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 8, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 8, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x7QI, true, RVVM1QI, 7, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x7QI, true, RVVMF2QI, 7, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x7QI, true, RVVMF4QI, 7, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 7, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x7QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 7, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x7QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 7, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 7, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x6QI, true, RVVM1QI, 6, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x6QI, true, RVVMF2QI, 6, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x6QI, true, RVVMF4QI, 6, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 6, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x6QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 6, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x6QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 6, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 6, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x5QI, true, RVVM1QI, 5, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x5QI, true, RVVMF2QI, 5, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x5QI, true, RVVMF4QI, 5, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 5, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x5QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 5, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x5QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 5, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 5, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x4QI, true, RVVM2QI, 4, LMUL_2, 4) TUPLE_ENTRY (RVVM1x4QI, true, RVVM1QI, 4, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x4QI, true, RVVMF2QI, 4, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x4QI, true, RVVMF4QI, 4, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 4, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x4QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 4, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x4QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 4, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 4, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x3QI, true, RVVM2QI, 3, LMUL_2, 4) TUPLE_ENTRY (RVVM1x3QI, true, RVVM1QI, 3, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x3QI, true, RVVMF2QI, 3, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x3QI, true, RVVMF4QI, 3, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 3, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x3QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 3, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x3QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 3, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 3, LMUL_F8, 64) TUPLE_ENTRY (RVVM4x2QI, true, RVVM4QI, 2, LMUL_4, 2) TUPLE_ENTRY (RVVM2x2QI, true, RVVM2QI, 2, LMUL_2, 4) TUPLE_ENTRY (RVVM1x2QI, true, RVVM1QI, 2, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x2QI, true, RVVMF2QI, 2, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x2QI, true, RVVMF4QI, 2, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 2, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x2QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 2, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x2QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 2, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 2, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x8HI, true, RVVM1HI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HI, true, RVVMF2HI, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HI, true, RVVM1HI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HI, true, RVVMF2HI, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HI, true, RVVM1HI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HI, true, RVVMF2HI, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HI, true, RVVM1HI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HI, true, RVVMF2HI, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HI, true, RVVM2HI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HI, true, RVVM1HI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HI, true, RVVMF2HI, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HI, true, RVVM2HI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HI, true, RVVM1HI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HI, true, RVVMF2HI, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HI, true, RVVM4HI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HI, true, RVVM2HI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HI, true, RVVM1HI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HI, true, RVVMF2HI, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM4HF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8SI, true, RVVM1SI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SI, true, RVVM1SI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SI, true, RVVM1SI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SI, true, RVVM1SI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SI, true, RVVM2SI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SI, true, RVVM1SI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SI, true, RVVM2SI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SI, true, RVVM1SI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SI, true, RVVM4SI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SI, true, RVVM2SI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SI, true, RVVM1SI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM4SF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 8, LMUL_1, 16) TUPLE_ENTRY (RVVM1x7DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 7, LMUL_1, 16) diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index 51878797287..44a36db7803 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -1406,6 +1406,9 @@ riscv_v_adjust_bytesize (machine_mode mode, int scale) { if (riscv_v_ext_vector_mode_p (mode)) { + if (TARGET_XTHEADVECTOR) + return BYTES_PER_RISCV_VECTOR; + poly_int64 nunits = GET_MODE_NUNITS (mode); poly_int64 mode_size = GET_MODE_SIZE (mode); @@ -9962,7 +9965,7 @@ riscv_use_divmod_expander (void) static machine_mode riscv_preferred_simd_mode (scalar_mode mode) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::preferred_simd_mode (mode); return word_mode; @@ -10313,7 +10316,7 @@ riscv_mode_priority (int, int n) unsigned int riscv_autovectorize_vector_modes (vector_modes *modes, bool all) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::autovectorize_vector_modes (modes, all); return default_autovectorize_vector_modes (modes, all); @@ -10496,6 +10499,16 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) return false; } +/* Implements target hook vector_mode_supported_any_target_p. */ + +static bool +riscv_vector_mode_supported_any_target_p (machine_mode mode) +{ + if (TARGET_XTHEADVECTOR) + return false; + return true; +} + /* Initialize the GCC target structure. */ #undef TARGET_ASM_ALIGNED_HI_OP #define TARGET_ASM_ALIGNED_HI_OP "\t.half\t" @@ -10839,6 +10852,9 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) #undef TARGET_PREFERRED_ELSE_VALUE #define TARGET_PREFERRED_ELSE_VALUE riscv_preferred_else_value +#undef TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P +#define TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P riscv_vector_mode_supported_any_target_p + struct gcc_target targetm = TARGET_INITIALIZER; #include "gt-riscv.h" diff --git a/gcc/config/riscv/riscv_th_vector.h b/gcc/config/riscv/riscv_th_vector.h new file mode 100644 index 00000000000..a897b9cb90a --- /dev/null +++ b/gcc/config/riscv/riscv_th_vector.h @@ -0,0 +1,49 @@ +/* RISC-V 'XTheadVector' Extension intrinsics include file. + Copyright (C) 2022-2023 Free Software Foundation, Inc. + + This file is part of GCC. + + GCC is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published + by the Free Software Foundation; either version 3, or (at your + option) any later version. + + GCC is distributed in the hope that it will be useful, but WITHOUT + ANY WARRANTY; without even the implied warranty of MERCHANTABILITY + or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public + License for more details. + + Under Section 7 of GPL version 3, you are granted additional + permissions described in the GCC Runtime Library Exception, version + 3.1, as published by the Free Software Foundation. + + You should have received a copy of the GNU General Public License and + a copy of the GCC Runtime Library Exception along with this program; + see the files COPYING3 and COPYING.RUNTIME respectively. If not, see + <" target="_blank">http://www.gnu.org/licenses/>. */ + +#ifndef __RISCV_TH_VECTOR_H +#define __RISCV_TH_VECTOR_H + +#include <stdint.h> +#include <stddef.h> + +#ifndef __riscv_xtheadvector +#error "XTheadVector intrinsics require the xtheadvector extension." +#else +#ifdef __cplusplus +extern "C" { +#endif + +/* NOTE: This implementation of riscv_th_vector.h is intentionally short. It does + not define the RVV types and intrinsic functions directly in C and C++ + code, but instead uses the following pragma to tell GCC to insert the + necessary type and function definitions itself. The net effect is the + same, and the file is a complete implementation of riscv_th_vector.h. */ +#pragma riscv intrinsic "xtheadvector" + +#ifdef __cplusplus +} +#endif // __cplusplus +#endif // __riscv_xtheadvector +#endif // __RISCV_TH_ECTOR_H diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md new file mode 100644 index 00000000000..1230bfb5994 --- /dev/null +++ b/gcc/config/riscv/thead-vector.md @@ -0,0 +1,69 @@ +(define_c_enum "unspec" [ + UNSPEC_TH_VWLDST +]) + +(define_mode_iterator V_VLS_VT [V VLS VT]) +(define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) + +(define_split + [(set (match_operand:V_VB_VLS_VT 0 "reg_or_mem_operand") + (match_operand:V_VB_VLS_VT 1 "reg_or_mem_operand"))] + "TARGET_XTHEADVECTOR" + [(const_int 0)] + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + }) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:V_VLS_VT 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:V_VLS_VT + [(match_operand:V_VLS_VT 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2")]) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:VB 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:VB + [(match_operand:VB 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2") + (set (attr "sew") (const_int 8)) + (set (attr "vlmul") (symbol_ref "riscv_vector::LMUL_1"))]) diff --git a/gcc/config/riscv/vector-iterators.md b/gcc/config/riscv/vector-iterators.md index 5f5f7b5b986..c0fc7a2441d 100644 --- a/gcc/config/riscv/vector-iterators.md +++ b/gcc/config/riscv/vector-iterators.md @@ -109,11 +109,11 @@ ]) (define_mode_iterator VI [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -128,11 +128,11 @@ ;; allow the instruction and mode to be matched during combine et al. (define_mode_iterator VF [ (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -140,11 +140,11 @@ (define_mode_iterator VF_ZVFHMIN [ (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -271,16 +271,16 @@ ]) (define_mode_iterator VEEWEXT2 [ - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -290,10 +290,10 @@ ]) (define_mode_iterator VEEWEXT4 [ - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -311,59 +311,59 @@ ]) (define_mode_iterator VEEWTRUNC2 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") (RVVM4SI "TARGET_64BIT") (RVVM2SI "TARGET_64BIT") (RVVM1SI "TARGET_64BIT") - (RVVMF2SI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_64BIT") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEEWTRUNC4 [ - RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM2HI "TARGET_64BIT") (RVVM1HI "TARGET_64BIT") - (RVVMF2HI "TARGET_64BIT") - (RVVMF4HI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2HI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") (RVVM1HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_64BIT") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEEWTRUNC8 [ (RVVM1QI "TARGET_64BIT") - (RVVMF2QI "TARGET_64BIT") - (RVVMF4QI "TARGET_64BIT") - (RVVMF8QI "TARGET_MIN_VLEN > 32 && TARGET_64BIT") + (RVVMF2QI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF4QI "!TARGET_XTHEADVECTOR && TARGET_64BIT") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32 && TARGET_64BIT") ]) (define_mode_iterator VEI16 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8HF "TARGET_VECTOR_ELEN_FP_16") (RVVM4HF "TARGET_VECTOR_ELEN_FP_16") (RVVM2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -452,11 +452,11 @@ ]) (define_mode_iterator VFULLI [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_FULL_V") (RVVM4DI "TARGET_FULL_V") (RVVM2DI "TARGET_FULL_V") (RVVM1DI "TARGET_FULL_V") @@ -509,17 +509,17 @@ ]) (define_mode_iterator VI_QH [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") ]) (define_mode_iterator VI_QHS [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -560,11 +560,11 @@ ]) (define_mode_iterator VI_QHS_NO_M8 [ - RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -603,11 +603,11 @@ (define_mode_iterator VF_HS [ (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") - (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (V1HF "riscv_vector::vls_mode_valid_p (V1HFmode) && TARGET_ZVFH") (V2HF "riscv_vector::vls_mode_valid_p (V2HFmode) && TARGET_ZVFH") @@ -638,12 +638,12 @@ (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") - (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (V1HF "riscv_vector::vls_mode_valid_p (V1HFmode) && TARGET_ZVFH") (V2HF "riscv_vector::vls_mode_valid_p (V2HFmode) && TARGET_ZVFH") @@ -674,11 +674,11 @@ ]) (define_mode_iterator V_VLSI_QHS [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (V1QI "riscv_vector::vls_mode_valid_p (V1QImode)") (V2QI "riscv_vector::vls_mode_valid_p (V2QImode)") @@ -756,27 +756,27 @@ ;; E.g. when index mode = RVVM8QImde and Pmode = SImode, if it is not zero_extend or ;; scalar != 1, such gather/scatter is not allowed since we don't have RVVM32SImode. (define_mode_iterator RATIO64 [ - (RVVMF8QI "TARGET_MIN_VLEN > 32") - (RVVMF4HI "TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM1DI "TARGET_VECTOR_ELEN_64") - (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") ]) (define_mode_iterator RATIO32 [ - RVVMF4QI - RVVMF2HI + (RVVMF4QI "!TARGET_XTHEADVECTOR") + (RVVMF2HI "!TARGET_XTHEADVECTOR") RVVM1SI (RVVM2DI "TARGET_VECTOR_ELEN_64") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") ]) (define_mode_iterator RATIO16 [ - RVVMF2QI + (RVVMF2QI "!TARGET_XTHEADVECTOR") RVVM1HI RVVM2SI (RVVM4DI "TARGET_VECTOR_ELEN_64") @@ -814,21 +814,21 @@ ]) (define_mode_iterator RATIO64I [ - (RVVMF8QI "TARGET_MIN_VLEN > 32") - (RVVMF4HI "TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM1DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") ]) (define_mode_iterator RATIO32I [ - RVVMF4QI - RVVMF2HI + (RVVMF4QI "!TARGET_XTHEADVECTOR") + (RVVMF2HI "!TARGET_XTHEADVECTOR") RVVM1SI (RVVM2DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") ]) (define_mode_iterator RATIO16I [ - RVVMF2QI + (RVVMF2QI "!TARGET_XTHEADVECTOR") RVVM1HI RVVM2SI (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") @@ -873,21 +873,21 @@ ]) (define_mode_iterator V_FRACT [ - RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - (RVVMF2HF "TARGET_VECTOR_ELEN_FP_16") (RVVMF4HF "TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16") (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32") - (RVVMF2SI "TARGET_MIN_VLEN > 32") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") ]) (define_mode_iterator VWEXTI [ - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -933,7 +933,7 @@ (RVVM4SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_16 && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -966,7 +966,7 @@ (RVVM4SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") @@ -996,7 +996,7 @@ (define_mode_iterator VWCONVERTI [ (RVVM8SI "TARGET_ZVFH") (RVVM4SI "TARGET_ZVFH") (RVVM2SI "TARGET_ZVFH") (RVVM1SI "TARGET_ZVFH") - (RVVMF2SI "TARGET_ZVFH") + (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_ZVFH") (RVVM8DI "TARGET_VECTOR_ELEN_64 && TARGET_VECTOR_ELEN_FP_32") (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_VECTOR_ELEN_FP_32") @@ -1045,7 +1045,7 @@ ]) (define_mode_iterator VQEXTI [ - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64") (RVVM4DI "TARGET_VECTOR_ELEN_64") (RVVM2DI "TARGET_VECTOR_ELEN_64") (RVVM1DI "TARGET_VECTOR_ELEN_64") @@ -1456,11 +1456,11 @@ ;; VINDEXED [VI8 VI16 VI32 (VI64 "TARGET_64BIT")]. (define_mode_iterator VINDEXED [ - RVVM8QI RVVM4QI RVVM2QI RVVM1QI RVVMF2QI RVVMF4QI (RVVMF8QI "TARGET_MIN_VLEN > 32") + RVVM8QI RVVM4QI RVVM2QI RVVM1QI (RVVMF2QI "!TARGET_XTHEADVECTOR") (RVVMF4QI "!TARGET_XTHEADVECTOR") (RVVMF8QI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8HI RVVM4HI RVVM2HI RVVM1HI RVVMF2HI (RVVMF4HI "TARGET_MIN_VLEN > 32") + RVVM8HI RVVM4HI RVVM2HI RVVM1HI (RVVMF2HI "!TARGET_XTHEADVECTOR") (RVVMF4HI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") - RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "TARGET_MIN_VLEN > 32") + RVVM8SI RVVM4SI RVVM2SI RVVM1SI (RVVMF2SI "!TARGET_XTHEADVECTOR && TARGET_MIN_VLEN > 32") (RVVM8DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") (RVVM4DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") @@ -1468,12 +1468,12 @@ (RVVM1DI "TARGET_VECTOR_ELEN_64 && TARGET_64BIT") (RVVM8HF "TARGET_ZVFH") (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") - (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64 && TARGET_64BIT") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64 && TARGET_64BIT") @@ -3173,11 +3173,11 @@ (define_mode_iterator V_VLS_F_CONVERT_SI [ (RVVM4HF "TARGET_ZVFH") (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") - (RVVMF2HF "TARGET_ZVFH") (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM8SF "TARGET_VECTOR_ELEN_FP_32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") @@ -3290,12 +3290,12 @@ ]) (define_mode_iterator V_VLS_F_CONVERT_DI [ - (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "TARGET_ZVFH") - (RVVMF4HF "TARGET_ZVFH && TARGET_MIN_VLEN > 32") + (RVVM2HF "TARGET_ZVFH") (RVVM1HF "TARGET_ZVFH") (RVVMF2HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH") + (RVVMF4HF "!TARGET_XTHEADVECTOR && TARGET_ZVFH && TARGET_MIN_VLEN > 32") (RVVM4SF "TARGET_VECTOR_ELEN_FP_32") (RVVM2SF "TARGET_VECTOR_ELEN_FP_32") (RVVM1SF "TARGET_VECTOR_ELEN_FP_32") - (RVVMF2SF "TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") + (RVVMF2SF "!TARGET_XTHEADVECTOR && TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32") (RVVM8DF "TARGET_VECTOR_ELEN_FP_64") (RVVM4DF "TARGET_VECTOR_ELEN_FP_64") (RVVM2DF "TARGET_VECTOR_ELEN_FP_64") (RVVM1DF "TARGET_VECTOR_ELEN_FP_64") diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index b5a9055cdc4..cb30c9ae97c 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -83,7 +83,7 @@ ;; check. However, we need default value of SEW for vsetvl instruction since there ;; is no field for ratio in the vsetvl instruction encoding. (define_attr "sew" "" - (cond [(eq_attr "mode" "RVVMF64BI,RVVMF32BI,RVVMF16BI,RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ + (cond [(eq_attr "mode" "RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ RVVM8QI,RVVM4QI,RVVM2QI,RVVM1QI,RVVMF2QI,RVVMF4QI,RVVMF8QI,\ RVVM1x8QI,RVVMF2x8QI,RVVMF4x8QI,RVVMF8x8QI,\ RVVM1x7QI,RVVMF2x7QI,RVVMF4x7QI,RVVMF8x7QI,\ @@ -95,6 +95,18 @@ V1QI,V2QI,V4QI,V8QI,V16QI,V32QI,V64QI,V128QI,V256QI,V512QI,V1024QI,V2048QI,V4096QI,\ V1BI,V2BI,V4BI,V8BI,V16BI,V32BI,V64BI,V128BI,V256BI,V512BI,V1024BI,V2048BI,V4096BI") (const_int 8) + (eq_attr "mode" "RVVMF16BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 16) + (const_int 8)) + (eq_attr "mode" "RVVMF32BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 32) + (const_int 8)) + (eq_attr "mode" "RVVMF64BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 64) + (const_int 8)) (eq_attr "mode" "RVVM8HI,RVVM4HI,RVVM2HI,RVVM1HI,RVVMF2HI,RVVMF4HI,\ RVVM1x8HI,RVVMF2x8HI,RVVMF4x8HI,\ RVVM1x7HI,RVVMF2x7HI,RVVMF4x7HI,\ @@ -155,9 +167,9 @@ (eq_attr "mode" "RVVM4QI,RVVMF2BI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2QI,RVVMF4BI") (symbol_ref "riscv_vector::LMUL_2") (eq_attr "mode" "RVVM1QI,RVVMF8BI") (symbol_ref "riscv_vector::LMUL_1") - (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "riscv_vector::LMUL_F2") - (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "riscv_vector::LMUL_F4") - (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "riscv_vector::LMUL_F8") + (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F2") + (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F4") + (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F8") (eq_attr "mode" "RVVM8HI") (symbol_ref "riscv_vector::LMUL_8") (eq_attr "mode" "RVVM4HI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2HI") (symbol_ref "riscv_vector::LMUL_2") @@ -428,6 +440,10 @@ vislide1up,vislide1down,vfslide1up,vfslide1down,\ vgather,vcompress,vlsegdux,vlsegdox,vssegtux,vssegtox") (const_int INVALID_ATTRIBUTE) + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE) (eq_attr "mode" "RVVM8QI,RVVM1BI") (const_int 1) (eq_attr "mode" "RVVM4QI,RVVMF2BI") (const_int 2) (eq_attr "mode" "RVVM2QI,RVVMF4BI") (const_int 4) @@ -888,6 +904,8 @@ (symbol_ref "riscv_vector::FRM_DYN")] (symbol_ref "riscv_vector::FRM_NONE"))) +(include "thead-vector.md") + ;; ----------------------------------------------------------------- ;; ---- Miscellaneous Operations ;; ----------------------------------------------------------------- @@ -1061,6 +1079,12 @@ - We can not leave it to TARGET_SECONDARY_RELOAD since it happens before spilling. The clobber scratch is used by spilling fractional registers in IRA/LRA so it's too early. */ + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; @@ -1097,7 +1121,7 @@ (define_insn "*mov<mode>_whole" [(set (match_operand:V_WHOLE 0 "reg_or_mem_operand" "=vr, m,vr") (match_operand:V_WHOLE 1 "reg_or_mem_operand" " m,vr,vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "@ vl%m1re<sew>.v\t%0,%1 vs%m1r.v\t%1,%0 @@ -1118,6 +1142,13 @@ (match_operand:VB 1 "general_operand"))] "TARGET_VECTOR" { + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } + if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; }) @@ -1125,7 +1156,7 @@ (define_insn "*mov<mode>" [(set (match_operand:VB 0 "register_operand" "=vr") (match_operand:VB 1 "register_operand" " vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "vmv1r.v\t%0,%1" [(set_attr "type" "vmov") (set_attr "mode" "<MODE>")]) @@ -1511,7 +1542,7 @@ (match_dup 4) (match_dup 5)] UNSPEC_VSETVL))] "TARGET_VECTOR" - "vset%i1vli\t%0,%1,e%2,%m3,t%p4,m%p5" + { return TARGET_XTHEADVECTOR ? "vsetvli\t%0,%1,e%2,%m3" : "vset%i1vli\t%0,%1,e%2,%m3,t%p4,m%p5"; } [(set_attr "type" "vsetvl") (set_attr "mode" "<MODE>") (set (attr "sew") (symbol_ref "INTVAL (operands[2])")) @@ -1529,7 +1560,7 @@ (match_operand 2 "const_int_operand" "i") (match_operand 3 "const_int_operand" "i")] UNSPEC_VSETVL))] "TARGET_VECTOR" - "vsetvli\tzero,zero,e%0,%m1,t%p2,m%p3" + { return TARGET_XTHEADVECTOR ? "vsetvli\tzero,zero,e%0,%m1" : "vsetvli\tzero,zero,e%0,%m1,t%p2,m%p3"; } [(set_attr "type" "vsetvl") (set_attr "mode" "SI") (set (attr "sew") (symbol_ref "INTVAL (operands[0])")) @@ -1551,7 +1582,7 @@ (match_operand 3 "const_int_operand" "i") (match_operand 4 "const_int_operand" "i")] UNSPEC_VSETVL))] "TARGET_VECTOR" - "vset%i0vli\tzero,%0,e%1,%m2,t%p3,m%p4" + { return TARGET_XTHEADVECTOR ? "vsetvli\tzero,%0,e%1,%m2" : "vset%i0vli\tzero,%0,e%1,%m2,t%p3,m%p4"; } [(set_attr "type" "vsetvl") (set_attr "mode" "<MODE>") (set (attr "sew") (symbol_ref "INTVAL (operands[1])")) @@ -3680,7 +3711,7 @@ (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3701,7 +3732,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3722,7 +3753,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c index 2e0e12aa045..2eef9e1e1a8 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c @@ -1,4 +1,4 @@ -/* { dg-do compile } */ +/* { dg-do compile { target { ! riscv_xtheadvector } } } */ /* { dg-skip-if "test rvv intrinsic" { *-*-* } { "*" } { "-march=rv*v*" } } */ void foo0 () {__rvv_bool64_t t;} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c index 3d81b179235..ef329e30785 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c @@ -1,4 +1,4 @@ /* { dg-do compile } */ /* { dg-options "-O3 -march=rv32gc -mabi=ilp32d" } */ -#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' extension enabled} } */ +#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' or 'XTHEADVECTOR' extension enabled} } */ diff --git a/gcc/testsuite/lib/target-supports.exp b/gcc/testsuite/lib/target-supports.exp index 7f13ff0ca56..70df6b1401c 100644 --- a/gcc/testsuite/lib/target-supports.exp +++ b/gcc/testsuite/lib/target-supports.exp @@ -1952,6 +1952,18 @@ proc check_effective_target_riscv_zbb { } { }] } +# Return 1 if the target arch supports the XTheadVector extension, 0 otherwise. +# Cache the result. + +proc check_effective_target_riscv_xtheadvector { } { + return [check_no_compiler_messages riscv_ext_xtheadvector assembly { + #ifndef __riscv_xtheadvector + #error "Not __riscv_xtheadvector" + #endif + }] +} + + # Return 1 if we can execute code when using dg-add-options riscv_v proc check_effective_target_riscv_v_ok { } { -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* [PATCH v4 6/6] RISC-V: Add support for xtheadvector-specific intrinsics. 2023-12-29 4:03 [PATCH v4] RISC-V: Support XTheadVector extension Jun Sha (Joshua) ` (4 preceding siblings ...) 2023-12-29 4:21 ` [PATCH v4] RISC-V: Handle differences between XTheadvector and Vector Jun Sha (Joshua) @ 2023-12-29 4:21 ` Jun Sha (Joshua) 2024-01-04 2:34 ` [PATCH v4] " Jun Sha (Joshua) 2024-01-03 2:37 ` [PATCH v4] RISC-V: Fix register overlap issue for some xtheadvector instructions Jun Sha (Joshua) ` (2 subsequent siblings) 8 siblings, 1 reply; 133+ messages in thread From: Jun Sha (Joshua) @ 2023-12-29 4:21 UTC (permalink / raw) To: gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, Jun Sha (Joshua), Jin Ma, Xianmiao Qu This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-functions.def (vlsegff): Include thead-vector-builtins-functions.def. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins-types.def (DEF_RVV_I8_OPS): Add datatypes for XTheadVector. (DEF_RVV_I16_OPS): Add datatypes for XTheadVector. (DEF_RVV_I32_OPS): Add datatypes for XTheadVector. (DEF_RVV_U8_OPS): Add datatypes for XTheadVector. (DEF_RVV_U16_OPS): Add datatypes for XTheadVector. (DEF_RVV_U32_OPS): Add datatypes for XTheadVector. (vint8m1_t): Add datatypes for XTheadVector. (vint8m2_t): Likewise. (vint8m4_t): Likewise. (vint8m8_t): Likewise. (vint16m1_t): Likewise. (vint16m2_t): Likewise. (vint16m4_t): Likewise. (vint16m8_t): Likewise. (vint32m1_t): Likewise. (vint32m2_t): Likewise. (vint32m4_t): Likewise. (vint32m8_t): Likewise. (vint64m1_t): Likewise. (vint64m2_t): Likewise. (vint64m4_t): Likewise. (vint64m8_t): Likewise. (vuint8m1_t): Likewise. (vuint8m2_t): Likewise. (vuint8m4_t): Likewise. (vuint8m8_t): Likewise. (vuint16m1_t): Likewise. (vuint16m2_t): Likewise. (vuint16m4_t): Likewise. (vuint16m8_t): Likewise. (vuint32m1_t): Likewise. (vuint32m2_t): Likewise. (vuint32m4_t): Likewise. (vuint32m8_t): Likewise. (vuint64m1_t): Likewise. (vuint64m2_t): Likewise. (vuint64m4_t): Likewise. (vuint64m8_t): Likewise. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_I8_OPS): Add datatypes for XTheadVector. (DEF_RVV_I16_OPS): Add datatypes for XTheadVector. (DEF_RVV_I32_OPS): Add datatypes for XTheadVector. (DEF_RVV_U8_OPS): Add datatypes for XTheadVector. (DEF_RVV_U16_OPS): Add datatypes for XTheadVector. (DEF_RVV_U32_OPS): Add datatypes for XTheadVector. * config/riscv/thead-vector-builtins-functions.def: New file. * config/riscv/thead-vector.md: Add new patterns. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config.gcc | 2 +- .../riscv/riscv-vector-builtins-shapes.cc | 126 +++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + .../riscv/riscv-vector-builtins-types.def | 120 +++++++ gcc/config/riscv/riscv-vector-builtins.cc | 313 +++++++++++++++++- gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 16 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector-builtins.cc | 200 +++++++++++ gcc/config/riscv/thead-vector-builtins.h | 64 ++++ gcc/config/riscv/thead-vector.md | 253 ++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 ++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 ++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 ++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 ++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 ++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 ++++ 17 files changed, 1545 insertions(+), 2 deletions(-) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/config/riscv/thead-vector-builtins.cc create mode 100644 gcc/config/riscv/thead-vector-builtins.h create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config.gcc b/gcc/config.gcc index 1445d98c147..4478395ab77 100644 --- a/gcc/config.gcc +++ b/gcc/config.gcc @@ -547,7 +547,7 @@ riscv*) extra_objs="riscv-builtins.o riscv-c.o riscv-sr.o riscv-shorten-memrefs.o riscv-selftests.o riscv-string.o" extra_objs="${extra_objs} riscv-v.o riscv-vsetvl.o riscv-vector-costs.o riscv-avlprop.o" extra_objs="${extra_objs} riscv-vector-builtins.o riscv-vector-builtins-shapes.o riscv-vector-builtins-bases.o" - extra_objs="${extra_objs} thead.o riscv-target-attr.o" + extra_objs="${extra_objs} thead.o riscv-target-attr.o thead-vector-builtins.o" d_target_objs="riscv-d.o" extra_headers="riscv_vector.h riscv_th_vector.h" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.cc" diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 6b49404a1fa..7d7c1f6f4b1 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,104 @@ struct indexed_loadstore_def : public function_shape } }; +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + /* Report an error if there is no xtheadvector. */ + if (!TARGET_XTHEADVECTOR) + return; + + build_all (b, group); + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Report an error if there is no xtheadvector. */ + if (!TARGET_XTHEADVECTOR) + return nullptr; + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_base_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + /* Report an error if there is no xtheadvector. */ + if (!TARGET_XTHEADVECTOR) + return; + + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_one (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_base_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +730,31 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + /* Report an error if there is no xtheadvector. */ + if (!TARGET_XTHEADVECTOR) + return; + + build_all (b, group); + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_base_name (instance.base_name); + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1011,6 +1134,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1023,6 +1148,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index df9884bb572..a822ba05bdd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins-types.def b/gcc/config/riscv/riscv-vector-builtins-types.def index 6aa45ae9a7e..e373d29e51c 100644 --- a/gcc/config/riscv/riscv-vector-builtins-types.def +++ b/gcc/config/riscv/riscv-vector-builtins-types.def @@ -24,12 +24,48 @@ along with GCC; see the file COPYING3. If not see #define DEF_RVV_I_OPS(TYPE, REQUIRE) #endif +/* Use "DEF_RVV_I8_OPS" macro include some signed integer (i8/i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I8_OPS +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_I16_OPS" macro include some signed integer (i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I16_OPS +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_I32_OPS" macro include some signed integer (i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I32_OPS +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) +#endif + /* Use "DEF_RVV_U_OPS" macro include all unsigned integer which will be iterated and registered as intrinsic functions. */ #ifndef DEF_RVV_U_OPS #define DEF_RVV_U_OPS(TYPE, REQUIRE) #endif +/* Use "DEF_RVV_U8_OPS" macro include some unsigned integer (i8/i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U8_OPS +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_U16_OPS" macro include some unsigned integer (i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U16_OPS +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_U32_OPS" macro include some unsigned integer (i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U32_OPS +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) +#endif + /* Use "DEF_RVV_F_OPS" macro include all floating-point which will be iterated and registered as intrinsic functions. */ #ifndef DEF_RVV_F_OPS @@ -362,6 +398,45 @@ DEF_RVV_I_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) DEF_RVV_I_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) DEF_RVV_I_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint8m1_t, 0) +DEF_RVV_I8_OPS (vint8m2_t, 0) +DEF_RVV_I8_OPS (vint8m4_t, 0) +DEF_RVV_I8_OPS (vint8m8_t, 0) +DEF_RVV_I8_OPS (vint16m1_t, 0) +DEF_RVV_I8_OPS (vint16m2_t, 0) +DEF_RVV_I8_OPS (vint16m4_t, 0) +DEF_RVV_I8_OPS (vint16m8_t, 0) +DEF_RVV_I8_OPS (vint32m1_t, 0) +DEF_RVV_I8_OPS (vint32m2_t, 0) +DEF_RVV_I8_OPS (vint32m4_t, 0) +DEF_RVV_I8_OPS (vint32m8_t, 0) +DEF_RVV_I8_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_I16_OPS (vint16m1_t, 0) +DEF_RVV_I16_OPS (vint16m2_t, 0) +DEF_RVV_I16_OPS (vint16m4_t, 0) +DEF_RVV_I16_OPS (vint16m8_t, 0) +DEF_RVV_I16_OPS (vint32m1_t, 0) +DEF_RVV_I16_OPS (vint32m2_t, 0) +DEF_RVV_I16_OPS (vint32m4_t, 0) +DEF_RVV_I16_OPS (vint32m8_t, 0) +DEF_RVV_I16_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_I32_OPS (vint32m1_t, 0) +DEF_RVV_I32_OPS (vint32m2_t, 0) +DEF_RVV_I32_OPS (vint32m4_t, 0) +DEF_RVV_I32_OPS (vint32m8_t, 0) +DEF_RVV_I32_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + DEF_RVV_U_OPS (vuint8mf8_t, RVV_REQUIRE_MIN_VLEN_64) DEF_RVV_U_OPS (vuint8mf4_t, 0) DEF_RVV_U_OPS (vuint8mf2_t, 0) @@ -385,6 +460,45 @@ DEF_RVV_U_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) DEF_RVV_U_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) DEF_RVV_U_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint8m1_t, 0) +DEF_RVV_U8_OPS (vuint8m2_t, 0) +DEF_RVV_U8_OPS (vuint8m4_t, 0) +DEF_RVV_U8_OPS (vuint8m8_t, 0) +DEF_RVV_U8_OPS (vuint16m1_t, 0) +DEF_RVV_U8_OPS (vuint16m2_t, 0) +DEF_RVV_U8_OPS (vuint16m4_t, 0) +DEF_RVV_U8_OPS (vuint16m8_t, 0) +DEF_RVV_U8_OPS (vuint32m1_t, 0) +DEF_RVV_U8_OPS (vuint32m2_t, 0) +DEF_RVV_U8_OPS (vuint32m4_t, 0) +DEF_RVV_U8_OPS (vuint32m8_t, 0) +DEF_RVV_U8_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_U16_OPS (vuint16m1_t, 0) +DEF_RVV_U16_OPS (vuint16m2_t, 0) +DEF_RVV_U16_OPS (vuint16m4_t, 0) +DEF_RVV_U16_OPS (vuint16m8_t, 0) +DEF_RVV_U16_OPS (vuint32m1_t, 0) +DEF_RVV_U16_OPS (vuint32m2_t, 0) +DEF_RVV_U16_OPS (vuint32m4_t, 0) +DEF_RVV_U16_OPS (vuint32m8_t, 0) +DEF_RVV_U16_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_U32_OPS (vuint32m1_t, 0) +DEF_RVV_U32_OPS (vuint32m2_t, 0) +DEF_RVV_U32_OPS (vuint32m4_t, 0) +DEF_RVV_U32_OPS (vuint32m8_t, 0) +DEF_RVV_U32_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + DEF_RVV_F_OPS (vfloat16mf4_t, RVV_REQUIRE_ELEN_FP_16 | RVV_REQUIRE_MIN_VLEN_64) DEF_RVV_F_OPS (vfloat16mf2_t, RVV_REQUIRE_ELEN_FP_16) DEF_RVV_F_OPS (vfloat16m1_t, RVV_REQUIRE_ELEN_FP_16) @@ -1356,7 +1470,13 @@ DEF_RVV_TUPLE_OPS (vfloat64m2x4_t, RVV_REQUIRE_ELEN_FP_64) DEF_RVV_TUPLE_OPS (vfloat64m4x2_t, RVV_REQUIRE_ELEN_FP_64) #undef DEF_RVV_I_OPS +#undef DEF_RVV_I8_OPS +#undef DEF_RVV_I16_OPS +#undef DEF_RVV_I32_OPS #undef DEF_RVV_U_OPS +#undef DEF_RVV_U8_OPS +#undef DEF_RVV_U16_OPS +#undef DEF_RVV_U32_OPS #undef DEF_RVV_F_OPS #undef DEF_RVV_B_OPS #undef DEF_RVV_WEXTI_OPS diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 4e2c66c2de7..461447afdef 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -51,6 +51,7 @@ #include "riscv-vector-builtins.h" #include "riscv-vector-builtins-shapes.h" #include "riscv-vector-builtins-bases.h" +#include "thead-vector-builtins.h" using namespace riscv_vector; @@ -246,6 +247,63 @@ static const rvv_type_info iu_ops[] = { #include "riscv-vector-builtins-types.def" {NUM_VECTOR_TYPES, 0}}; +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i8_ops[] = { +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i16_ops[] = { +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i32_ops[] = { +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u8_ops[] = { +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u16_ops[] = { +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u32_ops[] = { +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu8_ops[] = { +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu16_ops[] = { +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu32_ops[] = { +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + /* A list of all types will be registered for intrinsic functions. */ static const rvv_type_info all_ops[] = { #define DEF_RVV_I_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, @@ -913,7 +971,32 @@ static CONSTEXPR const rvv_arg_type_info tuple_vcreate_args[] /* A list of args for vector_type func (vector_type) function. */ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] - = {rvv_arg_type_info (RVV_BASE_vector), + = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; /* A list of none preds that will be registered for intrinsic functions. */ @@ -1429,6 +1512,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2604,6 +2695,222 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_size_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_size_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_size_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_size_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_size_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_size_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_index_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_index_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_index_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_index_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_index_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_index_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew8_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_index_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew16_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_index_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew32_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_index_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type, + * function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type) + * function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type) + * function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_size_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_size_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_size_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + /* A list of all RVV base function types. */ static CONSTEXPR const function_type_info function_types[] = { #define DEF_RVV_TYPE_INDEX( \ @@ -2687,6 +2994,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 4f38c09d73d..234b6f7a196 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -123,6 +123,7 @@ enum required_ext ZVKNHB_EXT, /* Crypto vector Zvknhb sub-ext */ ZVKSED_EXT, /* Crypto vector Zvksed sub-ext */ ZVKSH_EXT, /* Crypto vector Zvksh sub-ext */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ }; /* Enumerates the RVV operand types. */ @@ -252,6 +253,8 @@ struct function_group_info return TARGET_ZVKSED; case ZVKSH_EXT: return TARGET_ZVKSH; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; default: gcc_unreachable (); } diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 067771e3c97..09512092056 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -23,6 +23,8 @@ riscv-vector-builtins.o: $(srcdir)/config/riscv/riscv-vector-builtins.cc \ $(srcdir)/config/riscv/riscv-vector-builtins-shapes.h \ $(srcdir)/config/riscv/riscv-vector-builtins-bases.h \ $(srcdir)/config/riscv/riscv-vector-builtins-types.def \ + $(srcdir)/config/riscv/thead-vector-builtins.h \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ $(RISCV_BUILTINS_H) $(COMPILER) -c $(ALL_COMPILERFLAGS) $(ALL_CPPFLAGS) $(INCLUDES) \ $(srcdir)/config/riscv/riscv-vector-builtins.cc @@ -50,6 +52,20 @@ riscv-vector-builtins-bases.o: \ $(COMPILER) -c $(ALL_COMPILERFLAGS) $(ALL_CPPFLAGS) $(INCLUDES) \ $(srcdir)/config/riscv/riscv-vector-builtins-bases.cc +thead-vector-builtins.o: \ + $(srcdir)/config/riscv/thead-vector-builtins.cc \ + $(CONFIG_H) $(SYSTEM_H) coretypes.h $(TM_H) $(TREE_H) $(RTL_H) \ + $(TM_P_H) memmodel.h insn-codes.h $(OPTABS_H) $(RECOG_H) \ + $(EXPR_H) $(BASIC_BLOCK_H) $(FUNCTION_H) fold-const.h $(GIMPLE_H) \ + gimple-iterator.h gimplify.h explow.h $(EMIT_RTL_H) tree-vector-builder.h \ + rtx-vector-builder.h \ + $(srcdir)/config/riscv/riscv-vector-builtins-shapes.h \ + $(srcdir)/config/riscv/riscv-vector-builtins-bases.h \ + $(srcdir)/config/riscv/thead-vector-builtins.h \ + $(RISCV_BUILTINS_H) + $(COMPILER) -c $(ALL_COMPILERFLAGS) $(ALL_CPPFLAGS) $(INCLUDES) \ + $(srcdir)/config/riscv/thead-vector-builtins.cc + riscv-sr.o: $(srcdir)/config/riscv/riscv-sr.cc $(CONFIG_H) \ $(SYSTEM_H) $(TM_H) $(COMPILER) -c $(ALL_COMPILERFLAGS) $(ALL_CPPFLAGS) $(INCLUDES) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..667820d4c3e --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (th_vlb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlbu, th_loadstore_width, full_preds, u8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlhu, th_loadstore_width, full_preds, u16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlwu, th_loadstore_width, full_preds, u32_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vsb, th_loadstore_width, none_m_preds, iu8_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vsh, th_loadstore_width, none_m_preds, iu16_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vsw, th_loadstore_width, none_m_preds, iu32_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vlsb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsbu, th_loadstore_width, full_preds, u8_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlshu, th_loadstore_width, full_preds, u16_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlswu, th_loadstore_width, full_preds, u32_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssb, th_loadstore_width, none_m_preds, iu8_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssh, th_loadstore_width, none_m_preds, iu16_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssw, th_loadstore_width, none_m_preds, iu32_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlxb, th_indexed_loadstore_width, full_preds, i8_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxh, th_indexed_loadstore_width, full_preds, i16_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxw, th_indexed_loadstore_width, full_preds, i32_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxbu, th_indexed_loadstore_width, full_preds, u8_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxhu, th_indexed_loadstore_width, full_preds, u16_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxwu, th_indexed_loadstore_width, full_preds, u32_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxb, th_indexed_loadstore_width, none_m_preds, iu8_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxh, th_indexed_loadstore_width, none_m_preds, iu16_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxw, th_indexed_loadstore_width, none_m_preds, iu32_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxb, th_indexed_loadstore_width, none_m_preds, iu8_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxh, th_indexed_loadstore_width, none_m_preds, iu16_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxw, th_indexed_loadstore_width, none_m_preds, iu32_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector-builtins.cc b/gcc/config/riscv/thead-vector-builtins.cc new file mode 100644 index 00000000000..c0002f255ee --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins.cc @@ -0,0 +1,200 @@ +/* function_base implementation for RISC-V XTheadVector Extension + for GNU compiler. + Copyright (C) 2022-2023 Free Software Foundation, Inc. + Contributed by Joshua (cooper.joshua@linux.alibaba.com), T-Head + Semiconductor Co., Ltd. + + This file is part of GCC. + + GCC is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published by + the Free Software Foundation; either version 3, or (at your option) + any later version. + + GCC is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + General Public License for more details. + + You should have received a copy of the GNU General Public License + along with GCC; see the file COPYING3. If not see + <http://www.gnu.org/licenses/>. */ + +#include "config.h" +#include "system.h" +#include "coretypes.h" +#include "tm.h" +#include "tree.h" +#include "rtl.h" +#include "tm_p.h" +#include "memmodel.h" +#include "insn-codes.h" +#include "optabs.h" +#include "recog.h" +#include "expr.h" +#include "basic-block.h" +#include "function.h" +#include "fold-const.h" +#include "gimple.h" +#include "gimple-iterator.h" +#include "gimplify.h" +#include "explow.h" +#include "emit-rtl.h" +#include "tree-vector-builder.h" +#include "rtx-vector-builder.h" +#include "riscv-vector-builtins.h" +#include "riscv-vector-builtins-shapes.h" +#include "riscv-vector-builtins-bases.h" +#include "thead-vector-builtins.h" + +using namespace riscv_vector; + +namespace riscv_vector { + +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> th_vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> th_vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> th_vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> th_vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> th_vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> th_vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> th_vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> th_vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> th_vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> th_vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> th_vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> th_vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> th_vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> th_vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> th_vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> th_vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> th_vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> th_vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> th_vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> th_vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> th_vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> th_vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> th_vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> th_vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> th_vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> th_vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> th_vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> th_vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> th_vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> th_vsuxw_obj; +static CONSTEXPR const th_extract th_vext_x_v_obj; + +/* Declare the function base NAME, pointing it to an instance + of class <NAME>_obj. */ +#define BASE(NAME) \ + namespace bases { const function_base *const NAME = &NAME##_obj; } + +BASE (th_vlb) +BASE (th_vlh) +BASE (th_vlw) +BASE (th_vlbu) +BASE (th_vlhu) +BASE (th_vlwu) +BASE (th_vsb) +BASE (th_vsh) +BASE (th_vsw) +BASE (th_vlsb) +BASE (th_vlsh) +BASE (th_vlsw) +BASE (th_vlsbu) +BASE (th_vlshu) +BASE (th_vlswu) +BASE (th_vssb) +BASE (th_vssh) +BASE (th_vssw) +BASE (th_vlxb) +BASE (th_vlxh) +BASE (th_vlxw) +BASE (th_vlxbu) +BASE (th_vlxhu) +BASE (th_vlxwu) +BASE (th_vsxb) +BASE (th_vsxh) +BASE (th_vsxw) +BASE (th_vsuxb) +BASE (th_vsuxh) +BASE (th_vsuxw) +BASE (th_vext_x_v) + +} // end namespace riscv_vector diff --git a/gcc/config/riscv/thead-vector-builtins.h b/gcc/config/riscv/thead-vector-builtins.h new file mode 100644 index 00000000000..4720c6334d8 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins.h @@ -0,0 +1,64 @@ +/* function_base declaration for RISC-V XTheadVector Extension + for GNU compiler. + Copyright (C) 2022-2023 Free Software Foundation, Inc. + Contributed by Joshua (cooper.joshua@linux.alibaba.com), T-Head + Semiconductor Co., Ltd. + + This file is part of GCC. + + GCC is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published by + the Free Software Foundation; either version 3, or (at your option) + any later version. + + GCC is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + General Public License for more details. + + You should have received a copy of the GNU General Public License + along with GCC; see the file COPYING3. If not see + <http://www.gnu.org/licenses/>. */ + +#ifndef GCC_THEAD_VECTOR_BUILTINS_H +#define GCC_THEAD_VECTOR_BUILTINS_H + +namespace riscv_vector { + +namespace bases { +extern const function_base *const th_vlb; +extern const function_base *const th_vlh; +extern const function_base *const th_vlw; +extern const function_base *const th_vlbu; +extern const function_base *const th_vlhu; +extern const function_base *const th_vlwu; +extern const function_base *const th_vsb; +extern const function_base *const th_vsh; +extern const function_base *const th_vsw; +extern const function_base *const th_vlsb; +extern const function_base *const th_vlsh; +extern const function_base *const th_vlsw; +extern const function_base *const th_vlsbu; +extern const function_base *const th_vlshu; +extern const function_base *const th_vlswu; +extern const function_base *const th_vssb; +extern const function_base *const th_vssh; +extern const function_base *const th_vssw; +extern const function_base *const th_vlxb; +extern const function_base *const th_vlxh; +extern const function_base *const th_vlxw; +extern const function_base *const th_vlxbu; +extern const function_base *const th_vlxhu; +extern const function_base *const th_vlxwu; +extern const function_base *const th_vsxb; +extern const function_base *const th_vsxh; +extern const function_base *const th_vsxw; +extern const function_base *const th_vsuxb; +extern const function_base *const th_vsuxh; +extern const function_base *const th_vsuxw; +extern const function_base *const th_vext_x_v; +} + +} // end namespace riscv_vector + +#endif diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index af77e2a8a9e..d653b944c36 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -140,3 +228,168 @@ "" [(set_attr "type" "vsetvl") (set_attr "mode" "SI")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..4e192bbf025 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..1538afec68e --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..bf4924a1d76 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..8c451845175 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..0f5b09684a5 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..aaa75be023d --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} \ No newline at end of file -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* [PATCH v4] RISC-V: Add support for xtheadvector-specific intrinsics. 2023-12-29 4:21 ` [PATCH v4 6/6] RISC-V: Add support for xtheadvector-specific intrinsics Jun Sha (Joshua) @ 2024-01-04 2:34 ` Jun Sha (Joshua) 2024-01-10 9:27 ` [PATCH v5] " Jun Sha (Joshua) 2024-01-10 9:31 ` [PATCH " Jun Sha (Joshua) 0 siblings, 2 replies; 133+ messages in thread From: Jun Sha (Joshua) @ 2024-01-04 2:34 UTC (permalink / raw) To: gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, Jun Sha (Joshua), Jin Ma, Xianmiao Qu This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-functions.def (vlsegff): Include thead-vector-builtins-functions.def. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins-types.def (DEF_RVV_I8_OPS): Add datatypes for XTheadVector. (DEF_RVV_I16_OPS): Add datatypes for XTheadVector. (DEF_RVV_I32_OPS): Add datatypes for XTheadVector. (DEF_RVV_U8_OPS): Add datatypes for XTheadVector. (DEF_RVV_U16_OPS): Add datatypes for XTheadVector. (DEF_RVV_U32_OPS): Add datatypes for XTheadVector. (vint8m1_t): Add datatypes for XTheadVector. (vint8m2_t): Likewise. (vint8m4_t): Likewise. (vint8m8_t): Likewise. (vint16m1_t): Likewise. (vint16m2_t): Likewise. (vint16m4_t): Likewise. (vint16m8_t): Likewise. (vint32m1_t): Likewise. (vint32m2_t): Likewise. (vint32m4_t): Likewise. (vint32m8_t): Likewise. (vint64m1_t): Likewise. (vint64m2_t): Likewise. (vint64m4_t): Likewise. (vint64m8_t): Likewise. (vuint8m1_t): Likewise. (vuint8m2_t): Likewise. (vuint8m4_t): Likewise. (vuint8m8_t): Likewise. (vuint16m1_t): Likewise. (vuint16m2_t): Likewise. (vuint16m4_t): Likewise. (vuint16m8_t): Likewise. (vuint32m1_t): Likewise. (vuint32m2_t): Likewise. (vuint32m4_t): Likewise. (vuint32m8_t): Likewise. (vuint64m1_t): Likewise. (vuint64m2_t): Likewise. (vuint64m4_t): Likewise. (vuint64m8_t): Likewise. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_I8_OPS): Add datatypes for XTheadVector. (DEF_RVV_I16_OPS): Add datatypes for XTheadVector. (DEF_RVV_I32_OPS): Add datatypes for XTheadVector. (DEF_RVV_U8_OPS): Add datatypes for XTheadVector. (DEF_RVV_U16_OPS): Add datatypes for XTheadVector. (DEF_RVV_U32_OPS): Add datatypes for XTheadVector. * config/riscv/thead-vector-builtins-functions.def: New file. * config/riscv/thead-vector.md: Add new patterns. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config.gcc | 2 +- .../riscv/riscv-vector-builtins-shapes.cc | 126 +++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + .../riscv/riscv-vector-builtins-types.def | 120 +++++++ gcc/config/riscv/riscv-vector-builtins.cc | 313 +++++++++++++++++- gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 16 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector-builtins.cc | 200 +++++++++++ gcc/config/riscv/thead-vector-builtins.h | 64 ++++ gcc/config/riscv/thead-vector.md | 253 ++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 ++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 ++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 ++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 ++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 ++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 ++++ 17 files changed, 1545 insertions(+), 2 deletions(-) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/config/riscv/thead-vector-builtins.cc create mode 100644 gcc/config/riscv/thead-vector-builtins.h create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config.gcc b/gcc/config.gcc index 1445d98c147..4478395ab77 100644 --- a/gcc/config.gcc +++ b/gcc/config.gcc @@ -547,7 +547,7 @@ riscv*) extra_objs="riscv-builtins.o riscv-c.o riscv-sr.o riscv-shorten-memrefs.o riscv-selftests.o riscv-string.o" extra_objs="${extra_objs} riscv-v.o riscv-vsetvl.o riscv-vector-costs.o riscv-avlprop.o" extra_objs="${extra_objs} riscv-vector-builtins.o riscv-vector-builtins-shapes.o riscv-vector-builtins-bases.o" - extra_objs="${extra_objs} thead.o riscv-target-attr.o" + extra_objs="${extra_objs} thead.o riscv-target-attr.o thead-vector-builtins.o" d_target_objs="riscv-d.o" extra_headers="riscv_vector.h riscv_th_vector.h" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.cc" diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 6b49404a1fa..7d7c1f6f4b1 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,104 @@ struct indexed_loadstore_def : public function_shape } }; +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + /* Report an error if there is no xtheadvector. */ + if (!TARGET_XTHEADVECTOR) + return; + + build_all (b, group); + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Report an error if there is no xtheadvector. */ + if (!TARGET_XTHEADVECTOR) + return nullptr; + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_base_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + /* Report an error if there is no xtheadvector. */ + if (!TARGET_XTHEADVECTOR) + return; + + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_one (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_base_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +730,31 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + /* Report an error if there is no xtheadvector. */ + if (!TARGET_XTHEADVECTOR) + return; + + build_all (b, group); + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_base_name (instance.base_name); + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1011,6 +1134,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1023,6 +1148,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index df9884bb572..a822ba05bdd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins-types.def b/gcc/config/riscv/riscv-vector-builtins-types.def index 6aa45ae9a7e..e373d29e51c 100644 --- a/gcc/config/riscv/riscv-vector-builtins-types.def +++ b/gcc/config/riscv/riscv-vector-builtins-types.def @@ -24,12 +24,48 @@ along with GCC; see the file COPYING3. If not see #define DEF_RVV_I_OPS(TYPE, REQUIRE) #endif +/* Use "DEF_RVV_I8_OPS" macro include some signed integer (i8/i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I8_OPS +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_I16_OPS" macro include some signed integer (i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I16_OPS +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_I32_OPS" macro include some signed integer (i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I32_OPS +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) +#endif + /* Use "DEF_RVV_U_OPS" macro include all unsigned integer which will be iterated and registered as intrinsic functions. */ #ifndef DEF_RVV_U_OPS #define DEF_RVV_U_OPS(TYPE, REQUIRE) #endif +/* Use "DEF_RVV_U8_OPS" macro include some unsigned integer (i8/i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U8_OPS +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_U16_OPS" macro include some unsigned integer (i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U16_OPS +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_U32_OPS" macro include some unsigned integer (i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U32_OPS +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) +#endif + /* Use "DEF_RVV_F_OPS" macro include all floating-point which will be iterated and registered as intrinsic functions. */ #ifndef DEF_RVV_F_OPS @@ -362,6 +398,45 @@ DEF_RVV_I_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) DEF_RVV_I_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) DEF_RVV_I_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint8m1_t, 0) +DEF_RVV_I8_OPS (vint8m2_t, 0) +DEF_RVV_I8_OPS (vint8m4_t, 0) +DEF_RVV_I8_OPS (vint8m8_t, 0) +DEF_RVV_I8_OPS (vint16m1_t, 0) +DEF_RVV_I8_OPS (vint16m2_t, 0) +DEF_RVV_I8_OPS (vint16m4_t, 0) +DEF_RVV_I8_OPS (vint16m8_t, 0) +DEF_RVV_I8_OPS (vint32m1_t, 0) +DEF_RVV_I8_OPS (vint32m2_t, 0) +DEF_RVV_I8_OPS (vint32m4_t, 0) +DEF_RVV_I8_OPS (vint32m8_t, 0) +DEF_RVV_I8_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_I16_OPS (vint16m1_t, 0) +DEF_RVV_I16_OPS (vint16m2_t, 0) +DEF_RVV_I16_OPS (vint16m4_t, 0) +DEF_RVV_I16_OPS (vint16m8_t, 0) +DEF_RVV_I16_OPS (vint32m1_t, 0) +DEF_RVV_I16_OPS (vint32m2_t, 0) +DEF_RVV_I16_OPS (vint32m4_t, 0) +DEF_RVV_I16_OPS (vint32m8_t, 0) +DEF_RVV_I16_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_I32_OPS (vint32m1_t, 0) +DEF_RVV_I32_OPS (vint32m2_t, 0) +DEF_RVV_I32_OPS (vint32m4_t, 0) +DEF_RVV_I32_OPS (vint32m8_t, 0) +DEF_RVV_I32_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + DEF_RVV_U_OPS (vuint8mf8_t, RVV_REQUIRE_MIN_VLEN_64) DEF_RVV_U_OPS (vuint8mf4_t, 0) DEF_RVV_U_OPS (vuint8mf2_t, 0) @@ -385,6 +460,45 @@ DEF_RVV_U_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) DEF_RVV_U_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) DEF_RVV_U_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint8m1_t, 0) +DEF_RVV_U8_OPS (vuint8m2_t, 0) +DEF_RVV_U8_OPS (vuint8m4_t, 0) +DEF_RVV_U8_OPS (vuint8m8_t, 0) +DEF_RVV_U8_OPS (vuint16m1_t, 0) +DEF_RVV_U8_OPS (vuint16m2_t, 0) +DEF_RVV_U8_OPS (vuint16m4_t, 0) +DEF_RVV_U8_OPS (vuint16m8_t, 0) +DEF_RVV_U8_OPS (vuint32m1_t, 0) +DEF_RVV_U8_OPS (vuint32m2_t, 0) +DEF_RVV_U8_OPS (vuint32m4_t, 0) +DEF_RVV_U8_OPS (vuint32m8_t, 0) +DEF_RVV_U8_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_U16_OPS (vuint16m1_t, 0) +DEF_RVV_U16_OPS (vuint16m2_t, 0) +DEF_RVV_U16_OPS (vuint16m4_t, 0) +DEF_RVV_U16_OPS (vuint16m8_t, 0) +DEF_RVV_U16_OPS (vuint32m1_t, 0) +DEF_RVV_U16_OPS (vuint32m2_t, 0) +DEF_RVV_U16_OPS (vuint32m4_t, 0) +DEF_RVV_U16_OPS (vuint32m8_t, 0) +DEF_RVV_U16_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_U32_OPS (vuint32m1_t, 0) +DEF_RVV_U32_OPS (vuint32m2_t, 0) +DEF_RVV_U32_OPS (vuint32m4_t, 0) +DEF_RVV_U32_OPS (vuint32m8_t, 0) +DEF_RVV_U32_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + DEF_RVV_F_OPS (vfloat16mf4_t, RVV_REQUIRE_ELEN_FP_16 | RVV_REQUIRE_MIN_VLEN_64) DEF_RVV_F_OPS (vfloat16mf2_t, RVV_REQUIRE_ELEN_FP_16) DEF_RVV_F_OPS (vfloat16m1_t, RVV_REQUIRE_ELEN_FP_16) @@ -1356,7 +1470,13 @@ DEF_RVV_TUPLE_OPS (vfloat64m2x4_t, RVV_REQUIRE_ELEN_FP_64) DEF_RVV_TUPLE_OPS (vfloat64m4x2_t, RVV_REQUIRE_ELEN_FP_64) #undef DEF_RVV_I_OPS +#undef DEF_RVV_I8_OPS +#undef DEF_RVV_I16_OPS +#undef DEF_RVV_I32_OPS #undef DEF_RVV_U_OPS +#undef DEF_RVV_U8_OPS +#undef DEF_RVV_U16_OPS +#undef DEF_RVV_U32_OPS #undef DEF_RVV_F_OPS #undef DEF_RVV_B_OPS #undef DEF_RVV_WEXTI_OPS diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 4e2c66c2de7..461447afdef 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -51,6 +51,7 @@ #include "riscv-vector-builtins.h" #include "riscv-vector-builtins-shapes.h" #include "riscv-vector-builtins-bases.h" +#include "thead-vector-builtins.h" using namespace riscv_vector; @@ -246,6 +247,63 @@ static const rvv_type_info iu_ops[] = { #include "riscv-vector-builtins-types.def" {NUM_VECTOR_TYPES, 0}}; +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i8_ops[] = { +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i16_ops[] = { +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i32_ops[] = { +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u8_ops[] = { +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u16_ops[] = { +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u32_ops[] = { +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu8_ops[] = { +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu16_ops[] = { +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu32_ops[] = { +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + /* A list of all types will be registered for intrinsic functions. */ static const rvv_type_info all_ops[] = { #define DEF_RVV_I_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, @@ -913,7 +971,32 @@ static CONSTEXPR const rvv_arg_type_info tuple_vcreate_args[] /* A list of args for vector_type func (vector_type) function. */ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] - = {rvv_arg_type_info (RVV_BASE_vector), + = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; /* A list of none preds that will be registered for intrinsic functions. */ @@ -1429,6 +1512,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2604,6 +2695,222 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_size_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_size_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_size_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_size_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_size_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_size_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_index_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_index_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_index_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_index_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_index_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_index_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew8_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_index_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew16_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_index_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew32_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_index_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type, + * function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type) + * function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type) + * function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_size_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_size_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_size_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + /* A list of all RVV base function types. */ static CONSTEXPR const function_type_info function_types[] = { #define DEF_RVV_TYPE_INDEX( \ @@ -2687,6 +2994,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 4f38c09d73d..234b6f7a196 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -123,6 +123,7 @@ enum required_ext ZVKNHB_EXT, /* Crypto vector Zvknhb sub-ext */ ZVKSED_EXT, /* Crypto vector Zvksed sub-ext */ ZVKSH_EXT, /* Crypto vector Zvksh sub-ext */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ }; /* Enumerates the RVV operand types. */ @@ -252,6 +253,8 @@ struct function_group_info return TARGET_ZVKSED; case ZVKSH_EXT: return TARGET_ZVKSH; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; default: gcc_unreachable (); } diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 067771e3c97..09512092056 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -23,6 +23,8 @@ riscv-vector-builtins.o: $(srcdir)/config/riscv/riscv-vector-builtins.cc \ $(srcdir)/config/riscv/riscv-vector-builtins-shapes.h \ $(srcdir)/config/riscv/riscv-vector-builtins-bases.h \ $(srcdir)/config/riscv/riscv-vector-builtins-types.def \ + $(srcdir)/config/riscv/thead-vector-builtins.h \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ $(RISCV_BUILTINS_H) $(COMPILER) -c $(ALL_COMPILERFLAGS) $(ALL_CPPFLAGS) $(INCLUDES) \ $(srcdir)/config/riscv/riscv-vector-builtins.cc @@ -50,6 +52,20 @@ riscv-vector-builtins-bases.o: \ $(COMPILER) -c $(ALL_COMPILERFLAGS) $(ALL_CPPFLAGS) $(INCLUDES) \ $(srcdir)/config/riscv/riscv-vector-builtins-bases.cc +thead-vector-builtins.o: \ + $(srcdir)/config/riscv/thead-vector-builtins.cc \ + $(CONFIG_H) $(SYSTEM_H) coretypes.h $(TM_H) $(TREE_H) $(RTL_H) \ + $(TM_P_H) memmodel.h insn-codes.h $(OPTABS_H) $(RECOG_H) \ + $(EXPR_H) $(BASIC_BLOCK_H) $(FUNCTION_H) fold-const.h $(GIMPLE_H) \ + gimple-iterator.h gimplify.h explow.h $(EMIT_RTL_H) tree-vector-builder.h \ + rtx-vector-builder.h \ + $(srcdir)/config/riscv/riscv-vector-builtins-shapes.h \ + $(srcdir)/config/riscv/riscv-vector-builtins-bases.h \ + $(srcdir)/config/riscv/thead-vector-builtins.h \ + $(RISCV_BUILTINS_H) + $(COMPILER) -c $(ALL_COMPILERFLAGS) $(ALL_CPPFLAGS) $(INCLUDES) \ + $(srcdir)/config/riscv/thead-vector-builtins.cc + riscv-sr.o: $(srcdir)/config/riscv/riscv-sr.cc $(CONFIG_H) \ $(SYSTEM_H) $(TM_H) $(COMPILER) -c $(ALL_COMPILERFLAGS) $(ALL_CPPFLAGS) $(INCLUDES) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..667820d4c3e --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (th_vlb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlbu, th_loadstore_width, full_preds, u8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlhu, th_loadstore_width, full_preds, u16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlwu, th_loadstore_width, full_preds, u32_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vsb, th_loadstore_width, none_m_preds, iu8_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vsh, th_loadstore_width, none_m_preds, iu16_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vsw, th_loadstore_width, none_m_preds, iu32_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vlsb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsbu, th_loadstore_width, full_preds, u8_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlshu, th_loadstore_width, full_preds, u16_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlswu, th_loadstore_width, full_preds, u32_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssb, th_loadstore_width, none_m_preds, iu8_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssh, th_loadstore_width, none_m_preds, iu16_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssw, th_loadstore_width, none_m_preds, iu32_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlxb, th_indexed_loadstore_width, full_preds, i8_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxh, th_indexed_loadstore_width, full_preds, i16_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxw, th_indexed_loadstore_width, full_preds, i32_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxbu, th_indexed_loadstore_width, full_preds, u8_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxhu, th_indexed_loadstore_width, full_preds, u16_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxwu, th_indexed_loadstore_width, full_preds, u32_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxb, th_indexed_loadstore_width, none_m_preds, iu8_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxh, th_indexed_loadstore_width, none_m_preds, iu16_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxw, th_indexed_loadstore_width, none_m_preds, iu32_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxb, th_indexed_loadstore_width, none_m_preds, iu8_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxh, th_indexed_loadstore_width, none_m_preds, iu16_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxw, th_indexed_loadstore_width, none_m_preds, iu32_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector-builtins.cc b/gcc/config/riscv/thead-vector-builtins.cc new file mode 100644 index 00000000000..c0002f255ee --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins.cc @@ -0,0 +1,200 @@ +/* function_base implementation for RISC-V XTheadVector Extension + for GNU compiler. + Copyright (C) 2022-2023 Free Software Foundation, Inc. + Contributed by Joshua (cooper.joshua@linux.alibaba.com), T-Head + Semiconductor Co., Ltd. + + This file is part of GCC. + + GCC is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published by + the Free Software Foundation; either version 3, or (at your option) + any later version. + + GCC is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + General Public License for more details. + + You should have received a copy of the GNU General Public License + along with GCC; see the file COPYING3. If not see + <http://www.gnu.org/licenses/>. */ + +#include "config.h" +#include "system.h" +#include "coretypes.h" +#include "tm.h" +#include "tree.h" +#include "rtl.h" +#include "tm_p.h" +#include "memmodel.h" +#include "insn-codes.h" +#include "optabs.h" +#include "recog.h" +#include "expr.h" +#include "basic-block.h" +#include "function.h" +#include "fold-const.h" +#include "gimple.h" +#include "gimple-iterator.h" +#include "gimplify.h" +#include "explow.h" +#include "emit-rtl.h" +#include "tree-vector-builder.h" +#include "rtx-vector-builder.h" +#include "riscv-vector-builtins.h" +#include "riscv-vector-builtins-shapes.h" +#include "riscv-vector-builtins-bases.h" +#include "thead-vector-builtins.h" + +using namespace riscv_vector; + +namespace riscv_vector { + +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> th_vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> th_vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> th_vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> th_vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> th_vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> th_vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> th_vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> th_vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> th_vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> th_vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> th_vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> th_vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> th_vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> th_vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> th_vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> th_vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> th_vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> th_vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> th_vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> th_vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> th_vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> th_vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> th_vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> th_vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> th_vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> th_vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> th_vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> th_vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> th_vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> th_vsuxw_obj; +static CONSTEXPR const th_extract th_vext_x_v_obj; + +/* Declare the function base NAME, pointing it to an instance + of class <NAME>_obj. */ +#define BASE(NAME) \ + namespace bases { const function_base *const NAME = &NAME##_obj; } + +BASE (th_vlb) +BASE (th_vlh) +BASE (th_vlw) +BASE (th_vlbu) +BASE (th_vlhu) +BASE (th_vlwu) +BASE (th_vsb) +BASE (th_vsh) +BASE (th_vsw) +BASE (th_vlsb) +BASE (th_vlsh) +BASE (th_vlsw) +BASE (th_vlsbu) +BASE (th_vlshu) +BASE (th_vlswu) +BASE (th_vssb) +BASE (th_vssh) +BASE (th_vssw) +BASE (th_vlxb) +BASE (th_vlxh) +BASE (th_vlxw) +BASE (th_vlxbu) +BASE (th_vlxhu) +BASE (th_vlxwu) +BASE (th_vsxb) +BASE (th_vsxh) +BASE (th_vsxw) +BASE (th_vsuxb) +BASE (th_vsuxh) +BASE (th_vsuxw) +BASE (th_vext_x_v) + +} // end namespace riscv_vector diff --git a/gcc/config/riscv/thead-vector-builtins.h b/gcc/config/riscv/thead-vector-builtins.h new file mode 100644 index 00000000000..4720c6334d8 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins.h @@ -0,0 +1,64 @@ +/* function_base declaration for RISC-V XTheadVector Extension + for GNU compiler. + Copyright (C) 2022-2023 Free Software Foundation, Inc. + Contributed by Joshua (cooper.joshua@linux.alibaba.com), T-Head + Semiconductor Co., Ltd. + + This file is part of GCC. + + GCC is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published by + the Free Software Foundation; either version 3, or (at your option) + any later version. + + GCC is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + General Public License for more details. + + You should have received a copy of the GNU General Public License + along with GCC; see the file COPYING3. If not see + <http://www.gnu.org/licenses/>. */ + +#ifndef GCC_THEAD_VECTOR_BUILTINS_H +#define GCC_THEAD_VECTOR_BUILTINS_H + +namespace riscv_vector { + +namespace bases { +extern const function_base *const th_vlb; +extern const function_base *const th_vlh; +extern const function_base *const th_vlw; +extern const function_base *const th_vlbu; +extern const function_base *const th_vlhu; +extern const function_base *const th_vlwu; +extern const function_base *const th_vsb; +extern const function_base *const th_vsh; +extern const function_base *const th_vsw; +extern const function_base *const th_vlsb; +extern const function_base *const th_vlsh; +extern const function_base *const th_vlsw; +extern const function_base *const th_vlsbu; +extern const function_base *const th_vlshu; +extern const function_base *const th_vlswu; +extern const function_base *const th_vssb; +extern const function_base *const th_vssh; +extern const function_base *const th_vssw; +extern const function_base *const th_vlxb; +extern const function_base *const th_vlxh; +extern const function_base *const th_vlxw; +extern const function_base *const th_vlxbu; +extern const function_base *const th_vlxhu; +extern const function_base *const th_vlxwu; +extern const function_base *const th_vsxb; +extern const function_base *const th_vsxh; +extern const function_base *const th_vsxw; +extern const function_base *const th_vsuxb; +extern const function_base *const th_vsuxh; +extern const function_base *const th_vsuxw; +extern const function_base *const th_vext_x_v; +} + +} // end namespace riscv_vector + +#endif diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..0f3700d9269 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,168 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..4e192bbf025 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..1538afec68e --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..bf4924a1d76 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..8c451845175 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..0f5b09684a5 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..aaa75be023d --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} \ No newline at end of file -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-04 2:34 ` [PATCH v4] " Jun Sha (Joshua) @ 2024-01-10 9:27 ` Jun Sha (Joshua) 2024-01-10 9:35 ` juzhe.zhong 2024-01-10 9:31 ` [PATCH " Jun Sha (Joshua) 1 sibling, 1 reply; 133+ messages in thread From: Jun Sha (Joshua) @ 2024-01-10 9:27 UTC (permalink / raw) To: gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, Jun Sha (Joshua), Jin Ma, Xianmiao Qu This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-functions.def (vlsegff): Include thead-vector-builtins-functions.def. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins-types.def (DEF_RVV_I8_OPS): Add datatypes for XTheadVector. (DEF_RVV_I16_OPS): Add datatypes for XTheadVector. (DEF_RVV_I32_OPS): Add datatypes for XTheadVector. (DEF_RVV_U8_OPS): Add datatypes for XTheadVector. (DEF_RVV_U16_OPS): Add datatypes for XTheadVector. (DEF_RVV_U32_OPS): Add datatypes for XTheadVector. (vint8m1_t): Add datatypes for XTheadVector. (vint8m2_t): Likewise. (vint8m4_t): Likewise. (vint8m8_t): Likewise. (vint16m1_t): Likewise. (vint16m2_t): Likewise. (vint16m4_t): Likewise. (vint16m8_t): Likewise. (vint32m1_t): Likewise. (vint32m2_t): Likewise. (vint32m4_t): Likewise. (vint32m8_t): Likewise. (vint64m1_t): Likewise. (vint64m2_t): Likewise. (vint64m4_t): Likewise. (vint64m8_t): Likewise. (vuint8m1_t): Likewise. (vuint8m2_t): Likewise. (vuint8m4_t): Likewise. (vuint8m8_t): Likewise. (vuint16m1_t): Likewise. (vuint16m2_t): Likewise. (vuint16m4_t): Likewise. (vuint16m8_t): Likewise. (vuint32m1_t): Likewise. (vuint32m2_t): Likewise. (vuint32m4_t): Likewise. (vuint32m8_t): Likewise. (vuint64m1_t): Likewise. (vuint64m2_t): Likewise. (vuint64m4_t): Likewise. (vuint64m8_t): Likewise. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_I8_OPS): Add datatypes for XTheadVector. (DEF_RVV_I16_OPS): Add datatypes for XTheadVector. (DEF_RVV_I32_OPS): Add datatypes for XTheadVector. (DEF_RVV_U8_OPS): Add datatypes for XTheadVector. (DEF_RVV_U16_OPS): Add datatypes for XTheadVector. (DEF_RVV_U32_OPS): Add datatypes for XTheadVector. * config/riscv/thead-vector-builtins-functions.def: New file. * config/riscv/thead-vector.md: Add new patterns. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 ++ .../riscv/riscv-vector-builtins-shapes.cc | 98 ++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + .../riscv/riscv-vector-builtins-types.def | 120 +++++++ gcc/config/riscv/riscv-vector-builtins.cc | 311 ++++++++++++++++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 253 ++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 ++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 ++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 ++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 ++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 ++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 ++++ 16 files changed, 1406 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 46f1a1da33e..5f44f31a12b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2125,6 +2125,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2587,6 +2664,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> th_vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> th_vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> th_vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> th_vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> th_vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> th_vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> th_vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> th_vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> th_vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> th_vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> th_vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> th_vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> th_vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> th_vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> th_vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> th_vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> th_vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> th_vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> th_vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> th_vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> th_vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> th_vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> th_vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> th_vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> th_vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> th_vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> th_vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> th_vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> th_vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> th_vsuxw_obj; +static CONSTEXPR const th_extract th_vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2878,6 +2986,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (th_vlb) +BASE (th_vlh) +BASE (th_vlw) +BASE (th_vlbu) +BASE (th_vlhu) +BASE (th_vlwu) +BASE (th_vsb) +BASE (th_vsh) +BASE (th_vsw) +BASE (th_vlsb) +BASE (th_vlsh) +BASE (th_vlsw) +BASE (th_vlsbu) +BASE (th_vlshu) +BASE (th_vlswu) +BASE (th_vssb) +BASE (th_vssh) +BASE (th_vssw) +BASE (th_vlxb) +BASE (th_vlxh) +BASE (th_vlxw) +BASE (th_vlxbu) +BASE (th_vlxhu) +BASE (th_vlxwu) +BASE (th_vsxb) +BASE (th_vsxh) +BASE (th_vsxw) +BASE (th_vsuxb) +BASE (th_vsuxh) +BASE (th_vsuxw) +BASE (th_vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 1122e3801a7..df43adf9a17 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -299,6 +299,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const th_vlb; +extern const function_base *const th_vlh; +extern const function_base *const th_vlw; +extern const function_base *const th_vlbu; +extern const function_base *const th_vlhu; +extern const function_base *const th_vlwu; +extern const function_base *const th_vsb; +extern const function_base *const th_vsh; +extern const function_base *const th_vsw; +extern const function_base *const th_vlsb; +extern const function_base *const th_vlsh; +extern const function_base *const th_vlsw; +extern const function_base *const th_vlsbu; +extern const function_base *const th_vlshu; +extern const function_base *const th_vlswu; +extern const function_base *const th_vssb; +extern const function_base *const th_vssh; +extern const function_base *const th_vssw; +extern const function_base *const th_vlxb; +extern const function_base *const th_vlxh; +extern const function_base *const th_vlxw; +extern const function_base *const th_vlxbu; +extern const function_base *const th_vlxhu; +extern const function_base *const th_vlxwu; +extern const function_base *const th_vsxb; +extern const function_base *const th_vsxh; +extern const function_base *const th_vsxw; +extern const function_base *const th_vsuxb; +extern const function_base *const th_vsuxh; +extern const function_base *const th_vsuxw; +extern const function_base *const th_vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..489a95cf684 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,86 @@ struct indexed_loadstore_def : public function_shape } }; +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_base_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_one (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_base_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +712,21 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_base_name (instance.base_name); + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1189,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1203,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins-types.def b/gcc/config/riscv/riscv-vector-builtins-types.def index 61019a56844..abfeb4fcd9b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-types.def +++ b/gcc/config/riscv/riscv-vector-builtins-types.def @@ -24,12 +24,48 @@ along with GCC; see the file COPYING3. If not see #define DEF_RVV_I_OPS(TYPE, REQUIRE) #endif +/* Use "DEF_RVV_I8_OPS" macro include some signed integer (i8/i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I8_OPS +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_I16_OPS" macro include some signed integer (i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I16_OPS +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_I32_OPS" macro include some signed integer (i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I32_OPS +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) +#endif + /* Use "DEF_RVV_U_OPS" macro include all unsigned integer which will be iterated and registered as intrinsic functions. */ #ifndef DEF_RVV_U_OPS #define DEF_RVV_U_OPS(TYPE, REQUIRE) #endif +/* Use "DEF_RVV_U8_OPS" macro include some unsigned integer (i8/i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U8_OPS +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_U16_OPS" macro include some unsigned integer (i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U16_OPS +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_U32_OPS" macro include some unsigned integer (i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U32_OPS +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) +#endif + /* Use "DEF_RVV_F_OPS" macro include all floating-point which will be iterated and registered as intrinsic functions. */ #ifndef DEF_RVV_F_OPS @@ -374,6 +410,45 @@ DEF_RVV_I_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) DEF_RVV_I_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) DEF_RVV_I_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint8m1_t, 0) +DEF_RVV_I8_OPS (vint8m2_t, 0) +DEF_RVV_I8_OPS (vint8m4_t, 0) +DEF_RVV_I8_OPS (vint8m8_t, 0) +DEF_RVV_I8_OPS (vint16m1_t, 0) +DEF_RVV_I8_OPS (vint16m2_t, 0) +DEF_RVV_I8_OPS (vint16m4_t, 0) +DEF_RVV_I8_OPS (vint16m8_t, 0) +DEF_RVV_I8_OPS (vint32m1_t, 0) +DEF_RVV_I8_OPS (vint32m2_t, 0) +DEF_RVV_I8_OPS (vint32m4_t, 0) +DEF_RVV_I8_OPS (vint32m8_t, 0) +DEF_RVV_I8_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_I16_OPS (vint16m1_t, 0) +DEF_RVV_I16_OPS (vint16m2_t, 0) +DEF_RVV_I16_OPS (vint16m4_t, 0) +DEF_RVV_I16_OPS (vint16m8_t, 0) +DEF_RVV_I16_OPS (vint32m1_t, 0) +DEF_RVV_I16_OPS (vint32m2_t, 0) +DEF_RVV_I16_OPS (vint32m4_t, 0) +DEF_RVV_I16_OPS (vint32m8_t, 0) +DEF_RVV_I16_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_I32_OPS (vint32m1_t, 0) +DEF_RVV_I32_OPS (vint32m2_t, 0) +DEF_RVV_I32_OPS (vint32m4_t, 0) +DEF_RVV_I32_OPS (vint32m8_t, 0) +DEF_RVV_I32_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + DEF_RVV_U_OPS (vuint8mf8_t, RVV_REQUIRE_MIN_VLEN_64) DEF_RVV_U_OPS (vuint8mf4_t, 0) DEF_RVV_U_OPS (vuint8mf2_t, 0) @@ -397,6 +472,45 @@ DEF_RVV_U_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) DEF_RVV_U_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) DEF_RVV_U_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint8m1_t, 0) +DEF_RVV_U8_OPS (vuint8m2_t, 0) +DEF_RVV_U8_OPS (vuint8m4_t, 0) +DEF_RVV_U8_OPS (vuint8m8_t, 0) +DEF_RVV_U8_OPS (vuint16m1_t, 0) +DEF_RVV_U8_OPS (vuint16m2_t, 0) +DEF_RVV_U8_OPS (vuint16m4_t, 0) +DEF_RVV_U8_OPS (vuint16m8_t, 0) +DEF_RVV_U8_OPS (vuint32m1_t, 0) +DEF_RVV_U8_OPS (vuint32m2_t, 0) +DEF_RVV_U8_OPS (vuint32m4_t, 0) +DEF_RVV_U8_OPS (vuint32m8_t, 0) +DEF_RVV_U8_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_U16_OPS (vuint16m1_t, 0) +DEF_RVV_U16_OPS (vuint16m2_t, 0) +DEF_RVV_U16_OPS (vuint16m4_t, 0) +DEF_RVV_U16_OPS (vuint16m8_t, 0) +DEF_RVV_U16_OPS (vuint32m1_t, 0) +DEF_RVV_U16_OPS (vuint32m2_t, 0) +DEF_RVV_U16_OPS (vuint32m4_t, 0) +DEF_RVV_U16_OPS (vuint32m8_t, 0) +DEF_RVV_U16_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_U32_OPS (vuint32m1_t, 0) +DEF_RVV_U32_OPS (vuint32m2_t, 0) +DEF_RVV_U32_OPS (vuint32m4_t, 0) +DEF_RVV_U32_OPS (vuint32m8_t, 0) +DEF_RVV_U32_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + DEF_RVV_F_OPS (vfloat16mf4_t, RVV_REQUIRE_ELEN_FP_16 | RVV_REQUIRE_MIN_VLEN_64) DEF_RVV_F_OPS (vfloat16mf2_t, RVV_REQUIRE_ELEN_FP_16) DEF_RVV_F_OPS (vfloat16m1_t, RVV_REQUIRE_ELEN_FP_16) @@ -1379,7 +1493,13 @@ DEF_RVV_CRYPTO_SEW64_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) DEF_RVV_CRYPTO_SEW64_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) #undef DEF_RVV_I_OPS +#undef DEF_RVV_I8_OPS +#undef DEF_RVV_I16_OPS +#undef DEF_RVV_I32_OPS #undef DEF_RVV_U_OPS +#undef DEF_RVV_U8_OPS +#undef DEF_RVV_U16_OPS +#undef DEF_RVV_U32_OPS #undef DEF_RVV_F_OPS #undef DEF_RVV_B_OPS #undef DEF_RVV_WEXTI_OPS diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..f429f12dc18 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -246,6 +246,63 @@ static const rvv_type_info iu_ops[] = { #include "riscv-vector-builtins-types.def" {NUM_VECTOR_TYPES, 0}}; +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i8_ops[] = { +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i16_ops[] = { +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i32_ops[] = { +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u8_ops[] = { +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u16_ops[] = { +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u32_ops[] = { +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu8_ops[] = { +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu16_ops[] = { +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu32_ops[] = { +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + /* A list of all types will be registered for intrinsic functions. */ static const rvv_type_info all_ops[] = { #define DEF_RVV_I_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, @@ -934,6 +991,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1538,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2729,222 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_size_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_size_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_size_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_size_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_size_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_size_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_index_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_index_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_index_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_index_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_index_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_index_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew8_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_index_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew16_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_index_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew32_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_index_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type, + * function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type) + * function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type) + * function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_size_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_size_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_size_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +3123,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..a8ee39a3cb2 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -114,6 +114,7 @@ static const unsigned int CP_WRITE_CSR = 1U << 5; enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ @@ -234,6 +235,8 @@ struct function_group_info { case VECTOR_EXT: return TARGET_VECTOR; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; case ZVBB_EXT: return TARGET_ZVBB; case ZVBB_OR_ZVKB_EXT: diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..667820d4c3e --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (th_vlb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlbu, th_loadstore_width, full_preds, u8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlhu, th_loadstore_width, full_preds, u16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlwu, th_loadstore_width, full_preds, u32_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vsb, th_loadstore_width, none_m_preds, iu8_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vsh, th_loadstore_width, none_m_preds, iu16_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vsw, th_loadstore_width, none_m_preds, iu32_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vlsb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsbu, th_loadstore_width, full_preds, u8_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlshu, th_loadstore_width, full_preds, u16_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlswu, th_loadstore_width, full_preds, u32_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssb, th_loadstore_width, none_m_preds, iu8_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssh, th_loadstore_width, none_m_preds, iu16_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssw, th_loadstore_width, none_m_preds, iu32_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlxb, th_indexed_loadstore_width, full_preds, i8_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxh, th_indexed_loadstore_width, full_preds, i16_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxw, th_indexed_loadstore_width, full_preds, i32_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxbu, th_indexed_loadstore_width, full_preds, u8_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxhu, th_indexed_loadstore_width, full_preds, u16_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxwu, th_indexed_loadstore_width, full_preds, u32_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxb, th_indexed_loadstore_width, none_m_preds, iu8_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxh, th_indexed_loadstore_width, none_m_preds, iu16_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxw, th_indexed_loadstore_width, none_m_preds, iu32_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxb, th_indexed_loadstore_width, none_m_preds, iu8_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxh, th_indexed_loadstore_width, none_m_preds, iu16_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxw, th_indexed_loadstore_width, none_m_preds, iu32_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..0f3700d9269 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,168 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..4e192bbf025 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..1538afec68e --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..bf4924a1d76 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..8c451845175 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..0f5b09684a5 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..aaa75be023d --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} \ No newline at end of file -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-10 9:27 ` [PATCH v5] " Jun Sha (Joshua) @ 2024-01-10 9:35 ` juzhe.zhong 2024-01-10 9:55 ` Re：[PATCH " joshua 0 siblings, 1 reply; 133+ messages in thread From: juzhe.zhong @ 2024-01-10 9:35 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, cooper.joshua, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 71732 bytes --] +DEF_RVV_FUNCTION (th_vlb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_ops) I think we should remove those many data structure you added like: i8_v_scalar_const_ptr_ops Instead, you should use all_v_scalar_const_ptr_ops And revise th_loadstore_width, append the name according TYPE_UNSIGNED and GET_MODE_BITSIZE (GET_MODE_INNER (TYPE_MODE (instance.op_info->args[i].get_tree_type (instance.type.index)))) juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-10 17:27 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-functions.def (vlsegff): Include thead-vector-builtins-functions.def. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins-types.def (DEF_RVV_I8_OPS): Add datatypes for XTheadVector. (DEF_RVV_I16_OPS): Add datatypes for XTheadVector. (DEF_RVV_I32_OPS): Add datatypes for XTheadVector. (DEF_RVV_U8_OPS): Add datatypes for XTheadVector. (DEF_RVV_U16_OPS): Add datatypes for XTheadVector. (DEF_RVV_U32_OPS): Add datatypes for XTheadVector. (vint8m1_t): Add datatypes for XTheadVector. (vint8m2_t): Likewise. (vint8m4_t): Likewise. (vint8m8_t): Likewise. (vint16m1_t): Likewise. (vint16m2_t): Likewise. (vint16m4_t): Likewise. (vint16m8_t): Likewise. (vint32m1_t): Likewise. (vint32m2_t): Likewise. (vint32m4_t): Likewise. (vint32m8_t): Likewise. (vint64m1_t): Likewise. (vint64m2_t): Likewise. (vint64m4_t): Likewise. (vint64m8_t): Likewise. (vuint8m1_t): Likewise. (vuint8m2_t): Likewise. (vuint8m4_t): Likewise. (vuint8m8_t): Likewise. (vuint16m1_t): Likewise. (vuint16m2_t): Likewise. (vuint16m4_t): Likewise. (vuint16m8_t): Likewise. (vuint32m1_t): Likewise. (vuint32m2_t): Likewise. (vuint32m4_t): Likewise. (vuint32m8_t): Likewise. (vuint64m1_t): Likewise. (vuint64m2_t): Likewise. (vuint64m4_t): Likewise. (vuint64m8_t): Likewise. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_I8_OPS): Add datatypes for XTheadVector. (DEF_RVV_I16_OPS): Add datatypes for XTheadVector. (DEF_RVV_I32_OPS): Add datatypes for XTheadVector. (DEF_RVV_U8_OPS): Add datatypes for XTheadVector. (DEF_RVV_U16_OPS): Add datatypes for XTheadVector. (DEF_RVV_U32_OPS): Add datatypes for XTheadVector. * config/riscv/thead-vector-builtins-functions.def: New file. * config/riscv/thead-vector.md: Add new patterns. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 ++ .../riscv/riscv-vector-builtins-shapes.cc | 98 ++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + .../riscv/riscv-vector-builtins-types.def | 120 +++++++ gcc/config/riscv/riscv-vector-builtins.cc | 311 ++++++++++++++++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 253 ++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 ++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 ++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 ++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 ++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 ++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 ++++ 16 files changed, 1406 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 46f1a1da33e..5f44f31a12b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2125,6 +2125,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2587,6 +2664,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> th_vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> th_vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> th_vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> th_vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> th_vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> th_vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> th_vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> th_vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> th_vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> th_vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> th_vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> th_vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> th_vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> th_vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> th_vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> th_vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> th_vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> th_vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> th_vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> th_vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> th_vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> th_vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> th_vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> th_vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> th_vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> th_vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> th_vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> th_vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> th_vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> th_vsuxw_obj; +static CONSTEXPR const th_extract th_vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2878,6 +2986,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (th_vlb) +BASE (th_vlh) +BASE (th_vlw) +BASE (th_vlbu) +BASE (th_vlhu) +BASE (th_vlwu) +BASE (th_vsb) +BASE (th_vsh) +BASE (th_vsw) +BASE (th_vlsb) +BASE (th_vlsh) +BASE (th_vlsw) +BASE (th_vlsbu) +BASE (th_vlshu) +BASE (th_vlswu) +BASE (th_vssb) +BASE (th_vssh) +BASE (th_vssw) +BASE (th_vlxb) +BASE (th_vlxh) +BASE (th_vlxw) +BASE (th_vlxbu) +BASE (th_vlxhu) +BASE (th_vlxwu) +BASE (th_vsxb) +BASE (th_vsxh) +BASE (th_vsxw) +BASE (th_vsuxb) +BASE (th_vsuxh) +BASE (th_vsuxw) +BASE (th_vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 1122e3801a7..df43adf9a17 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -299,6 +299,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const th_vlb; +extern const function_base *const th_vlh; +extern const function_base *const th_vlw; +extern const function_base *const th_vlbu; +extern const function_base *const th_vlhu; +extern const function_base *const th_vlwu; +extern const function_base *const th_vsb; +extern const function_base *const th_vsh; +extern const function_base *const th_vsw; +extern const function_base *const th_vlsb; +extern const function_base *const th_vlsh; +extern const function_base *const th_vlsw; +extern const function_base *const th_vlsbu; +extern const function_base *const th_vlshu; +extern const function_base *const th_vlswu; +extern const function_base *const th_vssb; +extern const function_base *const th_vssh; +extern const function_base *const th_vssw; +extern const function_base *const th_vlxb; +extern const function_base *const th_vlxh; +extern const function_base *const th_vlxw; +extern const function_base *const th_vlxbu; +extern const function_base *const th_vlxhu; +extern const function_base *const th_vlxwu; +extern const function_base *const th_vsxb; +extern const function_base *const th_vsxh; +extern const function_base *const th_vsxw; +extern const function_base *const th_vsuxb; +extern const function_base *const th_vsuxh; +extern const function_base *const th_vsuxw; +extern const function_base *const th_vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..489a95cf684 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,86 @@ struct indexed_loadstore_def : public function_shape } }; +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_base_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_one (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_base_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +712,21 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_base_name (instance.base_name); + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1189,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1203,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins-types.def b/gcc/config/riscv/riscv-vector-builtins-types.def index 61019a56844..abfeb4fcd9b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-types.def +++ b/gcc/config/riscv/riscv-vector-builtins-types.def @@ -24,12 +24,48 @@ along with GCC; see the file COPYING3. If not see #define DEF_RVV_I_OPS(TYPE, REQUIRE) #endif +/* Use "DEF_RVV_I8_OPS" macro include some signed integer (i8/i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I8_OPS +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_I16_OPS" macro include some signed integer (i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I16_OPS +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_I32_OPS" macro include some signed integer (i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I32_OPS +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) +#endif + /* Use "DEF_RVV_U_OPS" macro include all unsigned integer which will be iterated and registered as intrinsic functions. */ #ifndef DEF_RVV_U_OPS #define DEF_RVV_U_OPS(TYPE, REQUIRE) #endif +/* Use "DEF_RVV_U8_OPS" macro include some unsigned integer (i8/i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U8_OPS +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_U16_OPS" macro include some unsigned integer (i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U16_OPS +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_U32_OPS" macro include some unsigned integer (i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U32_OPS +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) +#endif + /* Use "DEF_RVV_F_OPS" macro include all floating-point which will be iterated and registered as intrinsic functions. */ #ifndef DEF_RVV_F_OPS @@ -374,6 +410,45 @@ DEF_RVV_I_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) DEF_RVV_I_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) DEF_RVV_I_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint8m1_t, 0) +DEF_RVV_I8_OPS (vint8m2_t, 0) +DEF_RVV_I8_OPS (vint8m4_t, 0) +DEF_RVV_I8_OPS (vint8m8_t, 0) +DEF_RVV_I8_OPS (vint16m1_t, 0) +DEF_RVV_I8_OPS (vint16m2_t, 0) +DEF_RVV_I8_OPS (vint16m4_t, 0) +DEF_RVV_I8_OPS (vint16m8_t, 0) +DEF_RVV_I8_OPS (vint32m1_t, 0) +DEF_RVV_I8_OPS (vint32m2_t, 0) +DEF_RVV_I8_OPS (vint32m4_t, 0) +DEF_RVV_I8_OPS (vint32m8_t, 0) +DEF_RVV_I8_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_I16_OPS (vint16m1_t, 0) +DEF_RVV_I16_OPS (vint16m2_t, 0) +DEF_RVV_I16_OPS (vint16m4_t, 0) +DEF_RVV_I16_OPS (vint16m8_t, 0) +DEF_RVV_I16_OPS (vint32m1_t, 0) +DEF_RVV_I16_OPS (vint32m2_t, 0) +DEF_RVV_I16_OPS (vint32m4_t, 0) +DEF_RVV_I16_OPS (vint32m8_t, 0) +DEF_RVV_I16_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_I32_OPS (vint32m1_t, 0) +DEF_RVV_I32_OPS (vint32m2_t, 0) +DEF_RVV_I32_OPS (vint32m4_t, 0) +DEF_RVV_I32_OPS (vint32m8_t, 0) +DEF_RVV_I32_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + DEF_RVV_U_OPS (vuint8mf8_t, RVV_REQUIRE_MIN_VLEN_64) DEF_RVV_U_OPS (vuint8mf4_t, 0) DEF_RVV_U_OPS (vuint8mf2_t, 0) @@ -397,6 +472,45 @@ DEF_RVV_U_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) DEF_RVV_U_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) DEF_RVV_U_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint8m1_t, 0) +DEF_RVV_U8_OPS (vuint8m2_t, 0) +DEF_RVV_U8_OPS (vuint8m4_t, 0) +DEF_RVV_U8_OPS (vuint8m8_t, 0) +DEF_RVV_U8_OPS (vuint16m1_t, 0) +DEF_RVV_U8_OPS (vuint16m2_t, 0) +DEF_RVV_U8_OPS (vuint16m4_t, 0) +DEF_RVV_U8_OPS (vuint16m8_t, 0) +DEF_RVV_U8_OPS (vuint32m1_t, 0) +DEF_RVV_U8_OPS (vuint32m2_t, 0) +DEF_RVV_U8_OPS (vuint32m4_t, 0) +DEF_RVV_U8_OPS (vuint32m8_t, 0) +DEF_RVV_U8_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_U16_OPS (vuint16m1_t, 0) +DEF_RVV_U16_OPS (vuint16m2_t, 0) +DEF_RVV_U16_OPS (vuint16m4_t, 0) +DEF_RVV_U16_OPS (vuint16m8_t, 0) +DEF_RVV_U16_OPS (vuint32m1_t, 0) +DEF_RVV_U16_OPS (vuint32m2_t, 0) +DEF_RVV_U16_OPS (vuint32m4_t, 0) +DEF_RVV_U16_OPS (vuint32m8_t, 0) +DEF_RVV_U16_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_U32_OPS (vuint32m1_t, 0) +DEF_RVV_U32_OPS (vuint32m2_t, 0) +DEF_RVV_U32_OPS (vuint32m4_t, 0) +DEF_RVV_U32_OPS (vuint32m8_t, 0) +DEF_RVV_U32_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + DEF_RVV_F_OPS (vfloat16mf4_t, RVV_REQUIRE_ELEN_FP_16 | RVV_REQUIRE_MIN_VLEN_64) DEF_RVV_F_OPS (vfloat16mf2_t, RVV_REQUIRE_ELEN_FP_16) DEF_RVV_F_OPS (vfloat16m1_t, RVV_REQUIRE_ELEN_FP_16) @@ -1379,7 +1493,13 @@ DEF_RVV_CRYPTO_SEW64_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) DEF_RVV_CRYPTO_SEW64_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) #undef DEF_RVV_I_OPS +#undef DEF_RVV_I8_OPS +#undef DEF_RVV_I16_OPS +#undef DEF_RVV_I32_OPS #undef DEF_RVV_U_OPS +#undef DEF_RVV_U8_OPS +#undef DEF_RVV_U16_OPS +#undef DEF_RVV_U32_OPS #undef DEF_RVV_F_OPS #undef DEF_RVV_B_OPS #undef DEF_RVV_WEXTI_OPS diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..f429f12dc18 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -246,6 +246,63 @@ static const rvv_type_info iu_ops[] = { #include "riscv-vector-builtins-types.def" {NUM_VECTOR_TYPES, 0}}; +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i8_ops[] = { +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i16_ops[] = { +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i32_ops[] = { +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u8_ops[] = { +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u16_ops[] = { +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u32_ops[] = { +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu8_ops[] = { +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu16_ops[] = { +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu32_ops[] = { +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + /* A list of all types will be registered for intrinsic functions. */ static const rvv_type_info all_ops[] = { #define DEF_RVV_I_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, @@ -934,6 +991,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1538,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2729,222 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_size_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_size_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_size_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_size_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_size_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_size_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_index_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_index_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_index_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_index_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_index_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_index_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew8_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_index_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew16_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_index_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew32_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_index_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type, + * function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type) + * function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type) + * function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_size_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_size_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_size_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +3123,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..a8ee39a3cb2 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -114,6 +114,7 @@ static const unsigned int CP_WRITE_CSR = 1U << 5; enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ @@ -234,6 +235,8 @@ struct function_group_info { case VECTOR_EXT: return TARGET_VECTOR; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; case ZVBB_EXT: return TARGET_ZVBB; case ZVBB_OR_ZVKB_EXT: diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..667820d4c3e --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (th_vlb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlbu, th_loadstore_width, full_preds, u8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlhu, th_loadstore_width, full_preds, u16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlwu, th_loadstore_width, full_preds, u32_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vsb, th_loadstore_width, none_m_preds, iu8_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vsh, th_loadstore_width, none_m_preds, iu16_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vsw, th_loadstore_width, none_m_preds, iu32_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vlsb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsbu, th_loadstore_width, full_preds, u8_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlshu, th_loadstore_width, full_preds, u16_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlswu, th_loadstore_width, full_preds, u32_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssb, th_loadstore_width, none_m_preds, iu8_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssh, th_loadstore_width, none_m_preds, iu16_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssw, th_loadstore_width, none_m_preds, iu32_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlxb, th_indexed_loadstore_width, full_preds, i8_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxh, th_indexed_loadstore_width, full_preds, i16_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxw, th_indexed_loadstore_width, full_preds, i32_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxbu, th_indexed_loadstore_width, full_preds, u8_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxhu, th_indexed_loadstore_width, full_preds, u16_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxwu, th_indexed_loadstore_width, full_preds, u32_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxb, th_indexed_loadstore_width, none_m_preds, iu8_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxh, th_indexed_loadstore_width, none_m_preds, iu16_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxw, th_indexed_loadstore_width, none_m_preds, iu32_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxb, th_indexed_loadstore_width, none_m_preds, iu8_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxh, th_indexed_loadstore_width, none_m_preds, iu16_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxw, th_indexed_loadstore_width, none_m_preds, iu32_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..0f3700d9269 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,168 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..4e192bbf025 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..1538afec68e --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..bf4924a1d76 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..8c451845175 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..0f5b09684a5 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..aaa75be023d --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} \ No newline at end of file -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-10 9:35 ` juzhe.zhong @ 2024-01-10 9:55 ` joshua 2024-01-10 10:03 ` juzhe.zhong 0 siblings, 1 reply; 133+ messages in thread From: joshua @ 2024-01-10 9:55 UTC (permalink / raw) To: juzhe.zhong, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu And revise th_loadstore_width, append the name according TYPE_UNSIGNED and GET_MODE_BITSIZE (GET_MODE_INNER (TYPE_MODE (instance.op_info->args[i].get_tree_type (instance.type.index)))) What do you mean by it? I'm a bit confused. Changing i8_v_scalar_const_ptr_ops into all_v_scalar_const_ptr_ops will expand the datatypes that can be used in th_vlb. Can we restrict again in th_loadstore_width? ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月10日(星期三) 17:35 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. +DEF_RVV_FUNCTION (th_vlb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_ops) I think we should remove those many data structure you added like: i8_v_scalar_const_ptr_ops Instead, you should use all_v_scalar_const_ptr_ops And revise th_loadstore_width, append the name according TYPE_UNSIGNED and GET_MODE_BITSIZE (GET_MODE_INNER (TYPE_MODE (instance.op_info->args[i].get_tree_type (instance.type.index)))) juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-10 17:27 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-functions.def (vlsegff): Include thead-vector-builtins-functions.def. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins-types.def (DEF_RVV_I8_OPS): Add datatypes for XTheadVector. (DEF_RVV_I16_OPS): Add datatypes for XTheadVector. (DEF_RVV_I32_OPS): Add datatypes for XTheadVector. (DEF_RVV_U8_OPS): Add datatypes for XTheadVector. (DEF_RVV_U16_OPS): Add datatypes for XTheadVector. (DEF_RVV_U32_OPS): Add datatypes for XTheadVector. (vint8m1_t): Add datatypes for XTheadVector. (vint8m2_t): Likewise. (vint8m4_t): Likewise. (vint8m8_t): Likewise. (vint16m1_t): Likewise. (vint16m2_t): Likewise. (vint16m4_t): Likewise. (vint16m8_t): Likewise. (vint32m1_t): Likewise. (vint32m2_t): Likewise. (vint32m4_t): Likewise. (vint32m8_t): Likewise. (vint64m1_t): Likewise. (vint64m2_t): Likewise. (vint64m4_t): Likewise. (vint64m8_t): Likewise. (vuint8m1_t): Likewise. (vuint8m2_t): Likewise. (vuint8m4_t): Likewise. (vuint8m8_t): Likewise. (vuint16m1_t): Likewise. (vuint16m2_t): Likewise. (vuint16m4_t): Likewise. (vuint16m8_t): Likewise. (vuint32m1_t): Likewise. (vuint32m2_t): Likewise. (vuint32m4_t): Likewise. (vuint32m8_t): Likewise. (vuint64m1_t): Likewise. (vuint64m2_t): Likewise. (vuint64m4_t): Likewise. (vuint64m8_t): Likewise. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_I8_OPS): Add datatypes for XTheadVector. (DEF_RVV_I16_OPS): Add datatypes for XTheadVector. (DEF_RVV_I32_OPS): Add datatypes for XTheadVector. (DEF_RVV_U8_OPS): Add datatypes for XTheadVector. (DEF_RVV_U16_OPS): Add datatypes for XTheadVector. (DEF_RVV_U32_OPS): Add datatypes for XTheadVector. * config/riscv/thead-vector-builtins-functions.def: New file. * config/riscv/thead-vector.md: Add new patterns. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 ++ .../riscv/riscv-vector-builtins-shapes.cc | 98 ++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + .../riscv/riscv-vector-builtins-types.def | 120 +++++++ gcc/config/riscv/riscv-vector-builtins.cc | 311 ++++++++++++++++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 253 ++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 ++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 ++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 ++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 ++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 ++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 ++++ 16 files changed, 1406 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 46f1a1da33e..5f44f31a12b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2125,6 +2125,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2587,6 +2664,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> th_vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> th_vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> th_vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> th_vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> th_vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> th_vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> th_vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> th_vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> th_vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> th_vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> th_vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> th_vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> th_vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> th_vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> th_vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> th_vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> th_vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> th_vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> th_vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> th_vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> th_vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> th_vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> th_vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> th_vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> th_vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> th_vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> th_vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> th_vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> th_vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> th_vsuxw_obj; +static CONSTEXPR const th_extract th_vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2878,6 +2986,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (th_vlb) +BASE (th_vlh) +BASE (th_vlw) +BASE (th_vlbu) +BASE (th_vlhu) +BASE (th_vlwu) +BASE (th_vsb) +BASE (th_vsh) +BASE (th_vsw) +BASE (th_vlsb) +BASE (th_vlsh) +BASE (th_vlsw) +BASE (th_vlsbu) +BASE (th_vlshu) +BASE (th_vlswu) +BASE (th_vssb) +BASE (th_vssh) +BASE (th_vssw) +BASE (th_vlxb) +BASE (th_vlxh) +BASE (th_vlxw) +BASE (th_vlxbu) +BASE (th_vlxhu) +BASE (th_vlxwu) +BASE (th_vsxb) +BASE (th_vsxh) +BASE (th_vsxw) +BASE (th_vsuxb) +BASE (th_vsuxh) +BASE (th_vsuxw) +BASE (th_vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 1122e3801a7..df43adf9a17 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -299,6 +299,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const th_vlb; +extern const function_base *const th_vlh; +extern const function_base *const th_vlw; +extern const function_base *const th_vlbu; +extern const function_base *const th_vlhu; +extern const function_base *const th_vlwu; +extern const function_base *const th_vsb; +extern const function_base *const th_vsh; +extern const function_base *const th_vsw; +extern const function_base *const th_vlsb; +extern const function_base *const th_vlsh; +extern const function_base *const th_vlsw; +extern const function_base *const th_vlsbu; +extern const function_base *const th_vlshu; +extern const function_base *const th_vlswu; +extern const function_base *const th_vssb; +extern const function_base *const th_vssh; +extern const function_base *const th_vssw; +extern const function_base *const th_vlxb; +extern const function_base *const th_vlxh; +extern const function_base *const th_vlxw; +extern const function_base *const th_vlxbu; +extern const function_base *const th_vlxhu; +extern const function_base *const th_vlxwu; +extern const function_base *const th_vsxb; +extern const function_base *const th_vsxh; +extern const function_base *const th_vsxw; +extern const function_base *const th_vsuxb; +extern const function_base *const th_vsuxh; +extern const function_base *const th_vsuxw; +extern const function_base *const th_vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..489a95cf684 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,86 @@ struct indexed_loadstore_def : public function_shape } }; +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_base_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_one (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_base_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +712,21 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_base_name (instance.base_name); + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1189,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1203,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins-types.def b/gcc/config/riscv/riscv-vector-builtins-types.def index 61019a56844..abfeb4fcd9b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-types.def +++ b/gcc/config/riscv/riscv-vector-builtins-types.def @@ -24,12 +24,48 @@ along with GCC; see the file COPYING3. If not see #define DEF_RVV_I_OPS(TYPE, REQUIRE) #endif +/* Use "DEF_RVV_I8_OPS" macro include some signed integer (i8/i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I8_OPS +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_I16_OPS" macro include some signed integer (i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I16_OPS +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_I32_OPS" macro include some signed integer (i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I32_OPS +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) +#endif + /* Use "DEF_RVV_U_OPS" macro include all unsigned integer which will be iterated and registered as intrinsic functions. */ #ifndef DEF_RVV_U_OPS #define DEF_RVV_U_OPS(TYPE, REQUIRE) #endif +/* Use "DEF_RVV_U8_OPS" macro include some unsigned integer (i8/i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U8_OPS +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_U16_OPS" macro include some unsigned integer (i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U16_OPS +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_U32_OPS" macro include some unsigned integer (i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U32_OPS +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) +#endif + /* Use "DEF_RVV_F_OPS" macro include all floating-point which will be iterated and registered as intrinsic functions. */ #ifndef DEF_RVV_F_OPS @@ -374,6 +410,45 @@ DEF_RVV_I_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) DEF_RVV_I_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) DEF_RVV_I_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint8m1_t, 0) +DEF_RVV_I8_OPS (vint8m2_t, 0) +DEF_RVV_I8_OPS (vint8m4_t, 0) +DEF_RVV_I8_OPS (vint8m8_t, 0) +DEF_RVV_I8_OPS (vint16m1_t, 0) +DEF_RVV_I8_OPS (vint16m2_t, 0) +DEF_RVV_I8_OPS (vint16m4_t, 0) +DEF_RVV_I8_OPS (vint16m8_t, 0) +DEF_RVV_I8_OPS (vint32m1_t, 0) +DEF_RVV_I8_OPS (vint32m2_t, 0) +DEF_RVV_I8_OPS (vint32m4_t, 0) +DEF_RVV_I8_OPS (vint32m8_t, 0) +DEF_RVV_I8_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_I16_OPS (vint16m1_t, 0) +DEF_RVV_I16_OPS (vint16m2_t, 0) +DEF_RVV_I16_OPS (vint16m4_t, 0) +DEF_RVV_I16_OPS (vint16m8_t, 0) +DEF_RVV_I16_OPS (vint32m1_t, 0) +DEF_RVV_I16_OPS (vint32m2_t, 0) +DEF_RVV_I16_OPS (vint32m4_t, 0) +DEF_RVV_I16_OPS (vint32m8_t, 0) +DEF_RVV_I16_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_I32_OPS (vint32m1_t, 0) +DEF_RVV_I32_OPS (vint32m2_t, 0) +DEF_RVV_I32_OPS (vint32m4_t, 0) +DEF_RVV_I32_OPS (vint32m8_t, 0) +DEF_RVV_I32_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + DEF_RVV_U_OPS (vuint8mf8_t, RVV_REQUIRE_MIN_VLEN_64) DEF_RVV_U_OPS (vuint8mf4_t, 0) DEF_RVV_U_OPS (vuint8mf2_t, 0) @@ -397,6 +472,45 @@ DEF_RVV_U_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) DEF_RVV_U_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) DEF_RVV_U_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint8m1_t, 0) +DEF_RVV_U8_OPS (vuint8m2_t, 0) +DEF_RVV_U8_OPS (vuint8m4_t, 0) +DEF_RVV_U8_OPS (vuint8m8_t, 0) +DEF_RVV_U8_OPS (vuint16m1_t, 0) +DEF_RVV_U8_OPS (vuint16m2_t, 0) +DEF_RVV_U8_OPS (vuint16m4_t, 0) +DEF_RVV_U8_OPS (vuint16m8_t, 0) +DEF_RVV_U8_OPS (vuint32m1_t, 0) +DEF_RVV_U8_OPS (vuint32m2_t, 0) +DEF_RVV_U8_OPS (vuint32m4_t, 0) +DEF_RVV_U8_OPS (vuint32m8_t, 0) +DEF_RVV_U8_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_U16_OPS (vuint16m1_t, 0) +DEF_RVV_U16_OPS (vuint16m2_t, 0) +DEF_RVV_U16_OPS (vuint16m4_t, 0) +DEF_RVV_U16_OPS (vuint16m8_t, 0) +DEF_RVV_U16_OPS (vuint32m1_t, 0) +DEF_RVV_U16_OPS (vuint32m2_t, 0) +DEF_RVV_U16_OPS (vuint32m4_t, 0) +DEF_RVV_U16_OPS (vuint32m8_t, 0) +DEF_RVV_U16_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_U32_OPS (vuint32m1_t, 0) +DEF_RVV_U32_OPS (vuint32m2_t, 0) +DEF_RVV_U32_OPS (vuint32m4_t, 0) +DEF_RVV_U32_OPS (vuint32m8_t, 0) +DEF_RVV_U32_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + DEF_RVV_F_OPS (vfloat16mf4_t, RVV_REQUIRE_ELEN_FP_16 | RVV_REQUIRE_MIN_VLEN_64) DEF_RVV_F_OPS (vfloat16mf2_t, RVV_REQUIRE_ELEN_FP_16) DEF_RVV_F_OPS (vfloat16m1_t, RVV_REQUIRE_ELEN_FP_16) @@ -1379,7 +1493,13 @@ DEF_RVV_CRYPTO_SEW64_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) DEF_RVV_CRYPTO_SEW64_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) #undef DEF_RVV_I_OPS +#undef DEF_RVV_I8_OPS +#undef DEF_RVV_I16_OPS +#undef DEF_RVV_I32_OPS #undef DEF_RVV_U_OPS +#undef DEF_RVV_U8_OPS +#undef DEF_RVV_U16_OPS +#undef DEF_RVV_U32_OPS #undef DEF_RVV_F_OPS #undef DEF_RVV_B_OPS #undef DEF_RVV_WEXTI_OPS diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..f429f12dc18 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -246,6 +246,63 @@ static const rvv_type_info iu_ops[] = { #include "riscv-vector-builtins-types.def" {NUM_VECTOR_TYPES, 0}}; +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i8_ops[] = { +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i16_ops[] = { +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i32_ops[] = { +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u8_ops[] = { +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u16_ops[] = { +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u32_ops[] = { +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu8_ops[] = { +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu16_ops[] = { +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu32_ops[] = { +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + /* A list of all types will be registered for intrinsic functions. */ static const rvv_type_info all_ops[] = { #define DEF_RVV_I_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, @@ -934,6 +991,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1538,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2729,222 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_size_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_size_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_size_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_size_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_size_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_size_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_index_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_index_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_index_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_index_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_index_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_index_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew8_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_index_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew16_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_index_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew32_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_index_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type, + * function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type) + * function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type) + * function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_size_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_size_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_size_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +3123,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..a8ee39a3cb2 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -114,6 +114,7 @@ static const unsigned int CP_WRITE_CSR = 1U << 5; enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ @@ -234,6 +235,8 @@ struct function_group_info { case VECTOR_EXT: return TARGET_VECTOR; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; case ZVBB_EXT: return TARGET_ZVBB; case ZVBB_OR_ZVKB_EXT: diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..667820d4c3e --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (th_vlb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlbu, th_loadstore_width, full_preds, u8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlhu, th_loadstore_width, full_preds, u16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlwu, th_loadstore_width, full_preds, u32_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vsb, th_loadstore_width, none_m_preds, iu8_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vsh, th_loadstore_width, none_m_preds, iu16_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vsw, th_loadstore_width, none_m_preds, iu32_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vlsb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsbu, th_loadstore_width, full_preds, u8_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlshu, th_loadstore_width, full_preds, u16_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlswu, th_loadstore_width, full_preds, u32_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssb, th_loadstore_width, none_m_preds, iu8_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssh, th_loadstore_width, none_m_preds, iu16_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssw, th_loadstore_width, none_m_preds, iu32_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlxb, th_indexed_loadstore_width, full_preds, i8_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxh, th_indexed_loadstore_width, full_preds, i16_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxw, th_indexed_loadstore_width, full_preds, i32_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxbu, th_indexed_loadstore_width, full_preds, u8_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxhu, th_indexed_loadstore_width, full_preds, u16_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxwu, th_indexed_loadstore_width, full_preds, u32_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxb, th_indexed_loadstore_width, none_m_preds, iu8_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxh, th_indexed_loadstore_width, none_m_preds, iu16_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxw, th_indexed_loadstore_width, none_m_preds, iu32_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxb, th_indexed_loadstore_width, none_m_preds, iu8_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxh, th_indexed_loadstore_width, none_m_preds, iu16_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxw, th_indexed_loadstore_width, none_m_preds, iu32_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..0f3700d9269 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,168 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..4e192bbf025 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..1538afec68e --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..bf4924a1d76 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..8c451845175 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..0f5b09684a5 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..aaa75be023d --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} \ No newline at end of file -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-10 9:55 ` Re：[PATCH " joshua @ 2024-01-10 10:03 ` juzhe.zhong 2024-01-10 10:57 ` Re：Re：[PATCH " joshua 0 siblings, 1 reply; 133+ messages in thread From: juzhe.zhong @ 2024-01-10 10:03 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 73910 bytes --] I mean change these: +DEF_RVV_FUNCTION (th_vlb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_ops) into a single: +DEF_RVV_FUNCTION (th_vl, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) and append "h", "w", or"b" according to TYPE_UNSIGNED and GET_MODE_BITSIZE (GET_MODE_INNER (TYPE_MODE (instance.op_info->args[i].get_tree_type (instance.type.index)))) in th_loadstore_width. It should definitely works, I allow this flexibility in design of the framework. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-10 17:55 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. And revise th_loadstore_width, append the name according TYPE_UNSIGNED and GET_MODE_BITSIZE (GET_MODE_INNER (TYPE_MODE (instance.op_info->args[i].get_tree_type (instance.type.index)))) What do you mean by it? I'm a bit confused. Changing i8_v_scalar_const_ptr_ops into all_v_scalar_const_ptr_ops will expand the datatypes that can be used in th_vlb. Can we restrict again in th_loadstore_width? ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月10日(星期三) 17:35 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. +DEF_RVV_FUNCTION (th_vlb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_ops) I think we should remove those many data structure you added like: i8_v_scalar_const_ptr_ops Instead, you should use all_v_scalar_const_ptr_ops And revise th_loadstore_width, append the name according TYPE_UNSIGNED and GET_MODE_BITSIZE (GET_MODE_INNER (TYPE_MODE (instance.op_info->args[i].get_tree_type (instance.type.index)))) juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-10 17:27 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-functions.def (vlsegff): Include thead-vector-builtins-functions.def. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins-types.def (DEF_RVV_I8_OPS): Add datatypes for XTheadVector. (DEF_RVV_I16_OPS): Add datatypes for XTheadVector. (DEF_RVV_I32_OPS): Add datatypes for XTheadVector. (DEF_RVV_U8_OPS): Add datatypes for XTheadVector. (DEF_RVV_U16_OPS): Add datatypes for XTheadVector. (DEF_RVV_U32_OPS): Add datatypes for XTheadVector. (vint8m1_t): Add datatypes for XTheadVector. (vint8m2_t): Likewise. (vint8m4_t): Likewise. (vint8m8_t): Likewise. (vint16m1_t): Likewise. (vint16m2_t): Likewise. (vint16m4_t): Likewise. (vint16m8_t): Likewise. (vint32m1_t): Likewise. (vint32m2_t): Likewise. (vint32m4_t): Likewise. (vint32m8_t): Likewise. (vint64m1_t): Likewise. (vint64m2_t): Likewise. (vint64m4_t): Likewise. (vint64m8_t): Likewise. (vuint8m1_t): Likewise. (vuint8m2_t): Likewise. (vuint8m4_t): Likewise. (vuint8m8_t): Likewise. (vuint16m1_t): Likewise. (vuint16m2_t): Likewise. (vuint16m4_t): Likewise. (vuint16m8_t): Likewise. (vuint32m1_t): Likewise. (vuint32m2_t): Likewise. (vuint32m4_t): Likewise. (vuint32m8_t): Likewise. (vuint64m1_t): Likewise. (vuint64m2_t): Likewise. (vuint64m4_t): Likewise. (vuint64m8_t): Likewise. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_I8_OPS): Add datatypes for XTheadVector. (DEF_RVV_I16_OPS): Add datatypes for XTheadVector. (DEF_RVV_I32_OPS): Add datatypes for XTheadVector. (DEF_RVV_U8_OPS): Add datatypes for XTheadVector. (DEF_RVV_U16_OPS): Add datatypes for XTheadVector. (DEF_RVV_U32_OPS): Add datatypes for XTheadVector. * config/riscv/thead-vector-builtins-functions.def: New file. * config/riscv/thead-vector.md: Add new patterns. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 ++ .../riscv/riscv-vector-builtins-shapes.cc | 98 ++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + .../riscv/riscv-vector-builtins-types.def | 120 +++++++ gcc/config/riscv/riscv-vector-builtins.cc | 311 ++++++++++++++++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 253 ++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 ++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 ++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 ++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 ++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 ++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 ++++ 16 files changed, 1406 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 46f1a1da33e..5f44f31a12b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2125,6 +2125,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2587,6 +2664,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> th_vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> th_vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> th_vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> th_vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> th_vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> th_vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> th_vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> th_vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> th_vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> th_vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> th_vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> th_vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> th_vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> th_vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> th_vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> th_vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> th_vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> th_vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> th_vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> th_vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> th_vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> th_vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> th_vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> th_vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> th_vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> th_vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> th_vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> th_vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> th_vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> th_vsuxw_obj; +static CONSTEXPR const th_extract th_vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2878,6 +2986,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (th_vlb) +BASE (th_vlh) +BASE (th_vlw) +BASE (th_vlbu) +BASE (th_vlhu) +BASE (th_vlwu) +BASE (th_vsb) +BASE (th_vsh) +BASE (th_vsw) +BASE (th_vlsb) +BASE (th_vlsh) +BASE (th_vlsw) +BASE (th_vlsbu) +BASE (th_vlshu) +BASE (th_vlswu) +BASE (th_vssb) +BASE (th_vssh) +BASE (th_vssw) +BASE (th_vlxb) +BASE (th_vlxh) +BASE (th_vlxw) +BASE (th_vlxbu) +BASE (th_vlxhu) +BASE (th_vlxwu) +BASE (th_vsxb) +BASE (th_vsxh) +BASE (th_vsxw) +BASE (th_vsuxb) +BASE (th_vsuxh) +BASE (th_vsuxw) +BASE (th_vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 1122e3801a7..df43adf9a17 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -299,6 +299,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const th_vlb; +extern const function_base *const th_vlh; +extern const function_base *const th_vlw; +extern const function_base *const th_vlbu; +extern const function_base *const th_vlhu; +extern const function_base *const th_vlwu; +extern const function_base *const th_vsb; +extern const function_base *const th_vsh; +extern const function_base *const th_vsw; +extern const function_base *const th_vlsb; +extern const function_base *const th_vlsh; +extern const function_base *const th_vlsw; +extern const function_base *const th_vlsbu; +extern const function_base *const th_vlshu; +extern const function_base *const th_vlswu; +extern const function_base *const th_vssb; +extern const function_base *const th_vssh; +extern const function_base *const th_vssw; +extern const function_base *const th_vlxb; +extern const function_base *const th_vlxh; +extern const function_base *const th_vlxw; +extern const function_base *const th_vlxbu; +extern const function_base *const th_vlxhu; +extern const function_base *const th_vlxwu; +extern const function_base *const th_vsxb; +extern const function_base *const th_vsxh; +extern const function_base *const th_vsxw; +extern const function_base *const th_vsuxb; +extern const function_base *const th_vsuxh; +extern const function_base *const th_vsuxw; +extern const function_base *const th_vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..489a95cf684 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,86 @@ struct indexed_loadstore_def : public function_shape } }; +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_base_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_one (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_base_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +712,21 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_base_name (instance.base_name); + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1189,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1203,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins-types.def b/gcc/config/riscv/riscv-vector-builtins-types.def index 61019a56844..abfeb4fcd9b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-types.def +++ b/gcc/config/riscv/riscv-vector-builtins-types.def @@ -24,12 +24,48 @@ along with GCC; see the file COPYING3. If not see #define DEF_RVV_I_OPS(TYPE, REQUIRE) #endif +/* Use "DEF_RVV_I8_OPS" macro include some signed integer (i8/i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I8_OPS +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_I16_OPS" macro include some signed integer (i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I16_OPS +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_I32_OPS" macro include some signed integer (i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I32_OPS +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) +#endif + /* Use "DEF_RVV_U_OPS" macro include all unsigned integer which will be iterated and registered as intrinsic functions. */ #ifndef DEF_RVV_U_OPS #define DEF_RVV_U_OPS(TYPE, REQUIRE) #endif +/* Use "DEF_RVV_U8_OPS" macro include some unsigned integer (i8/i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U8_OPS +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_U16_OPS" macro include some unsigned integer (i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U16_OPS +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_U32_OPS" macro include some unsigned integer (i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U32_OPS +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) +#endif + /* Use "DEF_RVV_F_OPS" macro include all floating-point which will be iterated and registered as intrinsic functions. */ #ifndef DEF_RVV_F_OPS @@ -374,6 +410,45 @@ DEF_RVV_I_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) DEF_RVV_I_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) DEF_RVV_I_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint8m1_t, 0) +DEF_RVV_I8_OPS (vint8m2_t, 0) +DEF_RVV_I8_OPS (vint8m4_t, 0) +DEF_RVV_I8_OPS (vint8m8_t, 0) +DEF_RVV_I8_OPS (vint16m1_t, 0) +DEF_RVV_I8_OPS (vint16m2_t, 0) +DEF_RVV_I8_OPS (vint16m4_t, 0) +DEF_RVV_I8_OPS (vint16m8_t, 0) +DEF_RVV_I8_OPS (vint32m1_t, 0) +DEF_RVV_I8_OPS (vint32m2_t, 0) +DEF_RVV_I8_OPS (vint32m4_t, 0) +DEF_RVV_I8_OPS (vint32m8_t, 0) +DEF_RVV_I8_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_I16_OPS (vint16m1_t, 0) +DEF_RVV_I16_OPS (vint16m2_t, 0) +DEF_RVV_I16_OPS (vint16m4_t, 0) +DEF_RVV_I16_OPS (vint16m8_t, 0) +DEF_RVV_I16_OPS (vint32m1_t, 0) +DEF_RVV_I16_OPS (vint32m2_t, 0) +DEF_RVV_I16_OPS (vint32m4_t, 0) +DEF_RVV_I16_OPS (vint32m8_t, 0) +DEF_RVV_I16_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_I32_OPS (vint32m1_t, 0) +DEF_RVV_I32_OPS (vint32m2_t, 0) +DEF_RVV_I32_OPS (vint32m4_t, 0) +DEF_RVV_I32_OPS (vint32m8_t, 0) +DEF_RVV_I32_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + DEF_RVV_U_OPS (vuint8mf8_t, RVV_REQUIRE_MIN_VLEN_64) DEF_RVV_U_OPS (vuint8mf4_t, 0) DEF_RVV_U_OPS (vuint8mf2_t, 0) @@ -397,6 +472,45 @@ DEF_RVV_U_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) DEF_RVV_U_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) DEF_RVV_U_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint8m1_t, 0) +DEF_RVV_U8_OPS (vuint8m2_t, 0) +DEF_RVV_U8_OPS (vuint8m4_t, 0) +DEF_RVV_U8_OPS (vuint8m8_t, 0) +DEF_RVV_U8_OPS (vuint16m1_t, 0) +DEF_RVV_U8_OPS (vuint16m2_t, 0) +DEF_RVV_U8_OPS (vuint16m4_t, 0) +DEF_RVV_U8_OPS (vuint16m8_t, 0) +DEF_RVV_U8_OPS (vuint32m1_t, 0) +DEF_RVV_U8_OPS (vuint32m2_t, 0) +DEF_RVV_U8_OPS (vuint32m4_t, 0) +DEF_RVV_U8_OPS (vuint32m8_t, 0) +DEF_RVV_U8_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_U16_OPS (vuint16m1_t, 0) +DEF_RVV_U16_OPS (vuint16m2_t, 0) +DEF_RVV_U16_OPS (vuint16m4_t, 0) +DEF_RVV_U16_OPS (vuint16m8_t, 0) +DEF_RVV_U16_OPS (vuint32m1_t, 0) +DEF_RVV_U16_OPS (vuint32m2_t, 0) +DEF_RVV_U16_OPS (vuint32m4_t, 0) +DEF_RVV_U16_OPS (vuint32m8_t, 0) +DEF_RVV_U16_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_U32_OPS (vuint32m1_t, 0) +DEF_RVV_U32_OPS (vuint32m2_t, 0) +DEF_RVV_U32_OPS (vuint32m4_t, 0) +DEF_RVV_U32_OPS (vuint32m8_t, 0) +DEF_RVV_U32_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + DEF_RVV_F_OPS (vfloat16mf4_t, RVV_REQUIRE_ELEN_FP_16 | RVV_REQUIRE_MIN_VLEN_64) DEF_RVV_F_OPS (vfloat16mf2_t, RVV_REQUIRE_ELEN_FP_16) DEF_RVV_F_OPS (vfloat16m1_t, RVV_REQUIRE_ELEN_FP_16) @@ -1379,7 +1493,13 @@ DEF_RVV_CRYPTO_SEW64_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) DEF_RVV_CRYPTO_SEW64_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) #undef DEF_RVV_I_OPS +#undef DEF_RVV_I8_OPS +#undef DEF_RVV_I16_OPS +#undef DEF_RVV_I32_OPS #undef DEF_RVV_U_OPS +#undef DEF_RVV_U8_OPS +#undef DEF_RVV_U16_OPS +#undef DEF_RVV_U32_OPS #undef DEF_RVV_F_OPS #undef DEF_RVV_B_OPS #undef DEF_RVV_WEXTI_OPS diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..f429f12dc18 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -246,6 +246,63 @@ static const rvv_type_info iu_ops[] = { #include "riscv-vector-builtins-types.def" {NUM_VECTOR_TYPES, 0}}; +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i8_ops[] = { +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i16_ops[] = { +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i32_ops[] = { +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u8_ops[] = { +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u16_ops[] = { +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u32_ops[] = { +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu8_ops[] = { +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu16_ops[] = { +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu32_ops[] = { +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + /* A list of all types will be registered for intrinsic functions. */ static const rvv_type_info all_ops[] = { #define DEF_RVV_I_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, @@ -934,6 +991,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1538,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2729,222 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_size_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_size_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_size_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_size_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_size_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_size_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_index_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_index_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_index_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_index_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_index_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_index_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew8_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_index_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew16_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_index_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew32_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_index_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type, + * function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type) + * function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type) + * function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_size_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_size_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_size_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +3123,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..a8ee39a3cb2 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -114,6 +114,7 @@ static const unsigned int CP_WRITE_CSR = 1U << 5; enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ @@ -234,6 +235,8 @@ struct function_group_info { case VECTOR_EXT: return TARGET_VECTOR; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; case ZVBB_EXT: return TARGET_ZVBB; case ZVBB_OR_ZVKB_EXT: diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..667820d4c3e --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (th_vlb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlbu, th_loadstore_width, full_preds, u8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlhu, th_loadstore_width, full_preds, u16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlwu, th_loadstore_width, full_preds, u32_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vsb, th_loadstore_width, none_m_preds, iu8_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vsh, th_loadstore_width, none_m_preds, iu16_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vsw, th_loadstore_width, none_m_preds, iu32_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vlsb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsbu, th_loadstore_width, full_preds, u8_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlshu, th_loadstore_width, full_preds, u16_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlswu, th_loadstore_width, full_preds, u32_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssb, th_loadstore_width, none_m_preds, iu8_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssh, th_loadstore_width, none_m_preds, iu16_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssw, th_loadstore_width, none_m_preds, iu32_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlxb, th_indexed_loadstore_width, full_preds, i8_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxh, th_indexed_loadstore_width, full_preds, i16_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxw, th_indexed_loadstore_width, full_preds, i32_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxbu, th_indexed_loadstore_width, full_preds, u8_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxhu, th_indexed_loadstore_width, full_preds, u16_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxwu, th_indexed_loadstore_width, full_preds, u32_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxb, th_indexed_loadstore_width, none_m_preds, iu8_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxh, th_indexed_loadstore_width, none_m_preds, iu16_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxw, th_indexed_loadstore_width, none_m_preds, iu32_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxb, th_indexed_loadstore_width, none_m_preds, iu8_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxh, th_indexed_loadstore_width, none_m_preds, iu16_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxw, th_indexed_loadstore_width, none_m_preds, iu32_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..0f3700d9269 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,168 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..4e192bbf025 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..1538afec68e --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..bf4924a1d76 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..8c451845175 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..0f5b09684a5 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..aaa75be023d --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} \ No newline at end of file -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-10 10:03 ` juzhe.zhong @ 2024-01-10 10:57 ` joshua 0 siblings, 0 replies; 133+ messages in thread From: joshua @ 2024-01-10 10:57 UTC (permalink / raw) To: juzhe.zhong, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu [-- Attachment #1.1: Type: text/plain, Size: 70720 bytes --] Hi Juzhe, Perhaps things are not as simple as imagined. The differences between vlb/vlh/vlw is not the same as vle8/vle16/vle32. "8", "16" or "32" in vle8/vle16/vle32 can be appended from "vle" according to input type. But vlb/vlh/vlw is different not in input type. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月10日(星期三) 18:03 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. I mean change these: +DEF_RVV_FUNCTION (th_vlb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_ops) into a single: +DEF_RVV_FUNCTION (th_vl, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) and append "h", "w", or"b" according to TYPE_UNSIGNED and GET_MODE_BITSIZE (GET_MODE_INNER (TYPE_MODE (instance.op_info->args[i].get_tree_type (instance.type.index)))) in th_loadstore_width. It should definitely works, I allow this flexibility in design of the framework. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-10 17:55 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. And revise th_loadstore_width, append the name according TYPE_UNSIGNED and GET_MODE_BITSIZE (GET_MODE_INNER (TYPE_MODE (instance.op_info->args[i].get_tree_type (instance.type.index)))) What do you mean by it? I'm a bit confused. Changing i8_v_scalar_const_ptr_ops into all_v_scalar_const_ptr_ops will expand the datatypes that can be used in th_vlb. Can we restrict again in th_loadstore_width? ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月10日(星期三) 17:35 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. +DEF_RVV_FUNCTION (th_vlb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_ops) I think we should remove those many data structure you added like: i8_v_scalar_const_ptr_ops Instead, you should use all_v_scalar_const_ptr_ops And revise th_loadstore_width, append the name according TYPE_UNSIGNED and GET_MODE_BITSIZE (GET_MODE_INNER (TYPE_MODE (instance.op_info->args[i].get_tree_type (instance.type.index)))) juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-10 17:27 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-functions.def (vlsegff): Include thead-vector-builtins-functions.def. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins-types.def (DEF_RVV_I8_OPS): Add datatypes for XTheadVector. (DEF_RVV_I16_OPS): Add datatypes for XTheadVector. (DEF_RVV_I32_OPS): Add datatypes for XTheadVector. (DEF_RVV_U8_OPS): Add datatypes for XTheadVector. (DEF_RVV_U16_OPS): Add datatypes for XTheadVector. (DEF_RVV_U32_OPS): Add datatypes for XTheadVector. (vint8m1_t): Add datatypes for XTheadVector. (vint8m2_t): Likewise. (vint8m4_t): Likewise. (vint8m8_t): Likewise. (vint16m1_t): Likewise. (vint16m2_t): Likewise. (vint16m4_t): Likewise. (vint16m8_t): Likewise. (vint32m1_t): Likewise. (vint32m2_t): Likewise. (vint32m4_t): Likewise. (vint32m8_t): Likewise. (vint64m1_t): Likewise. (vint64m2_t): Likewise. (vint64m4_t): Likewise. (vint64m8_t): Likewise. (vuint8m1_t): Likewise. (vuint8m2_t): Likewise. (vuint8m4_t): Likewise. (vuint8m8_t): Likewise. (vuint16m1_t): Likewise. (vuint16m2_t): Likewise. (vuint16m4_t): Likewise. (vuint16m8_t): Likewise. (vuint32m1_t): Likewise. (vuint32m2_t): Likewise. (vuint32m4_t): Likewise. (vuint32m8_t): Likewise. (vuint64m1_t): Likewise. (vuint64m2_t): Likewise. (vuint64m4_t): Likewise. (vuint64m8_t): Likewise. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_I8_OPS): Add datatypes for XTheadVector. (DEF_RVV_I16_OPS): Add datatypes for XTheadVector. (DEF_RVV_I32_OPS): Add datatypes for XTheadVector. (DEF_RVV_U8_OPS): Add datatypes for XTheadVector. (DEF_RVV_U16_OPS): Add datatypes for XTheadVector. (DEF_RVV_U32_OPS): Add datatypes for XTheadVector. * config/riscv/thead-vector-builtins-functions.def: New file. * config/riscv/thead-vector.md: Add new patterns. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 ++ .../riscv/riscv-vector-builtins-shapes.cc | 98 ++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + .../riscv/riscv-vector-builtins-types.def | 120 +++++++ gcc/config/riscv/riscv-vector-builtins.cc | 311 ++++++++++++++++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 253 ++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 ++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 ++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 ++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 ++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 ++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 ++++ 16 files changed, 1406 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 46f1a1da33e..5f44f31a12b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2125,6 +2125,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2587,6 +2664,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> th_vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> th_vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> th_vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> th_vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> th_vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> th_vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> th_vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> th_vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> th_vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> th_vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> th_vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> th_vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> th_vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> th_vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> th_vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> th_vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> th_vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> th_vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> th_vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> th_vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> th_vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> th_vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> th_vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> th_vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> th_vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> th_vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> th_vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> th_vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> th_vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> th_vsuxw_obj; +static CONSTEXPR const th_extract th_vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2878,6 +2986,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (th_vlb) +BASE (th_vlh) +BASE (th_vlw) +BASE (th_vlbu) +BASE (th_vlhu) +BASE (th_vlwu) +BASE (th_vsb) +BASE (th_vsh) +BASE (th_vsw) +BASE (th_vlsb) +BASE (th_vlsh) +BASE (th_vlsw) +BASE (th_vlsbu) +BASE (th_vlshu) +BASE (th_vlswu) +BASE (th_vssb) +BASE (th_vssh) +BASE (th_vssw) +BASE (th_vlxb) +BASE (th_vlxh) +BASE (th_vlxw) +BASE (th_vlxbu) +BASE (th_vlxhu) +BASE (th_vlxwu) +BASE (th_vsxb) +BASE (th_vsxh) +BASE (th_vsxw) +BASE (th_vsuxb) +BASE (th_vsuxh) +BASE (th_vsuxw) +BASE (th_vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 1122e3801a7..df43adf9a17 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -299,6 +299,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const th_vlb; +extern const function_base *const th_vlh; +extern const function_base *const th_vlw; +extern const function_base *const th_vlbu; +extern const function_base *const th_vlhu; +extern const function_base *const th_vlwu; +extern const function_base *const th_vsb; +extern const function_base *const th_vsh; +extern const function_base *const th_vsw; +extern const function_base *const th_vlsb; +extern const function_base *const th_vlsh; +extern const function_base *const th_vlsw; +extern const function_base *const th_vlsbu; +extern const function_base *const th_vlshu; +extern const function_base *const th_vlswu; +extern const function_base *const th_vssb; +extern const function_base *const th_vssh; +extern const function_base *const th_vssw; +extern const function_base *const th_vlxb; +extern const function_base *const th_vlxh; +extern const function_base *const th_vlxw; +extern const function_base *const th_vlxbu; +extern const function_base *const th_vlxhu; +extern const function_base *const th_vlxwu; +extern const function_base *const th_vsxb; +extern const function_base *const th_vsxh; +extern const function_base *const th_vsxw; +extern const function_base *const th_vsuxb; +extern const function_base *const th_vsuxh; +extern const function_base *const th_vsuxw; +extern const function_base *const th_vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..489a95cf684 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,86 @@ struct indexed_loadstore_def : public function_shape } }; +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_base_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_one (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_base_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +712,21 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_base_name (instance.base_name); + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1189,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1203,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins-types.def b/gcc/config/riscv/riscv-vector-builtins-types.def index 61019a56844..abfeb4fcd9b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-types.def +++ b/gcc/config/riscv/riscv-vector-builtins-types.def @@ -24,12 +24,48 @@ along with GCC; see the file COPYING3. If not see #define DEF_RVV_I_OPS(TYPE, REQUIRE) #endif +/* Use "DEF_RVV_I8_OPS" macro include some signed integer (i8/i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I8_OPS +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_I16_OPS" macro include some signed integer (i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I16_OPS +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_I32_OPS" macro include some signed integer (i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I32_OPS +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) +#endif + /* Use "DEF_RVV_U_OPS" macro include all unsigned integer which will be iterated and registered as intrinsic functions. */ #ifndef DEF_RVV_U_OPS #define DEF_RVV_U_OPS(TYPE, REQUIRE) #endif +/* Use "DEF_RVV_U8_OPS" macro include some unsigned integer (i8/i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U8_OPS +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_U16_OPS" macro include some unsigned integer (i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U16_OPS +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_U32_OPS" macro include some unsigned integer (i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U32_OPS +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) +#endif + /* Use "DEF_RVV_F_OPS" macro include all floating-point which will be iterated and registered as intrinsic functions. */ #ifndef DEF_RVV_F_OPS @@ -374,6 +410,45 @@ DEF_RVV_I_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) DEF_RVV_I_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) DEF_RVV_I_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint8m1_t, 0) +DEF_RVV_I8_OPS (vint8m2_t, 0) +DEF_RVV_I8_OPS (vint8m4_t, 0) +DEF_RVV_I8_OPS (vint8m8_t, 0) +DEF_RVV_I8_OPS (vint16m1_t, 0) +DEF_RVV_I8_OPS (vint16m2_t, 0) +DEF_RVV_I8_OPS (vint16m4_t, 0) +DEF_RVV_I8_OPS (vint16m8_t, 0) +DEF_RVV_I8_OPS (vint32m1_t, 0) +DEF_RVV_I8_OPS (vint32m2_t, 0) +DEF_RVV_I8_OPS (vint32m4_t, 0) +DEF_RVV_I8_OPS (vint32m8_t, 0) +DEF_RVV_I8_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_I16_OPS (vint16m1_t, 0) +DEF_RVV_I16_OPS (vint16m2_t, 0) +DEF_RVV_I16_OPS (vint16m4_t, 0) +DEF_RVV_I16_OPS (vint16m8_t, 0) +DEF_RVV_I16_OPS (vint32m1_t, 0) +DEF_RVV_I16_OPS (vint32m2_t, 0) +DEF_RVV_I16_OPS (vint32m4_t, 0) +DEF_RVV_I16_OPS (vint32m8_t, 0) +DEF_RVV_I16_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_I32_OPS (vint32m1_t, 0) +DEF_RVV_I32_OPS (vint32m2_t, 0) +DEF_RVV_I32_OPS (vint32m4_t, 0) +DEF_RVV_I32_OPS (vint32m8_t, 0) +DEF_RVV_I32_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + DEF_RVV_U_OPS (vuint8mf8_t, RVV_REQUIRE_MIN_VLEN_64) DEF_RVV_U_OPS (vuint8mf4_t, 0) DEF_RVV_U_OPS (vuint8mf2_t, 0) @@ -397,6 +472,45 @@ DEF_RVV_U_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) DEF_RVV_U_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) DEF_RVV_U_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint8m1_t, 0) +DEF_RVV_U8_OPS (vuint8m2_t, 0) +DEF_RVV_U8_OPS (vuint8m4_t, 0) +DEF_RVV_U8_OPS (vuint8m8_t, 0) +DEF_RVV_U8_OPS (vuint16m1_t, 0) +DEF_RVV_U8_OPS (vuint16m2_t, 0) +DEF_RVV_U8_OPS (vuint16m4_t, 0) +DEF_RVV_U8_OPS (vuint16m8_t, 0) +DEF_RVV_U8_OPS (vuint32m1_t, 0) +DEF_RVV_U8_OPS (vuint32m2_t, 0) +DEF_RVV_U8_OPS (vuint32m4_t, 0) +DEF_RVV_U8_OPS (vuint32m8_t, 0) +DEF_RVV_U8_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_U16_OPS (vuint16m1_t, 0) +DEF_RVV_U16_OPS (vuint16m2_t, 0) +DEF_RVV_U16_OPS (vuint16m4_t, 0) +DEF_RVV_U16_OPS (vuint16m8_t, 0) +DEF_RVV_U16_OPS (vuint32m1_t, 0) +DEF_RVV_U16_OPS (vuint32m2_t, 0) +DEF_RVV_U16_OPS (vuint32m4_t, 0) +DEF_RVV_U16_OPS (vuint32m8_t, 0) +DEF_RVV_U16_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_U32_OPS (vuint32m1_t, 0) +DEF_RVV_U32_OPS (vuint32m2_t, 0) +DEF_RVV_U32_OPS (vuint32m4_t, 0) +DEF_RVV_U32_OPS (vuint32m8_t, 0) +DEF_RVV_U32_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + DEF_RVV_F_OPS (vfloat16mf4_t, RVV_REQUIRE_ELEN_FP_16 | RVV_REQUIRE_MIN_VLEN_64) DEF_RVV_F_OPS (vfloat16mf2_t, RVV_REQUIRE_ELEN_FP_16) DEF_RVV_F_OPS (vfloat16m1_t, RVV_REQUIRE_ELEN_FP_16) @@ -1379,7 +1493,13 @@ DEF_RVV_CRYPTO_SEW64_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) DEF_RVV_CRYPTO_SEW64_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) #undef DEF_RVV_I_OPS +#undef DEF_RVV_I8_OPS +#undef DEF_RVV_I16_OPS +#undef DEF_RVV_I32_OPS #undef DEF_RVV_U_OPS +#undef DEF_RVV_U8_OPS +#undef DEF_RVV_U16_OPS +#undef DEF_RVV_U32_OPS #undef DEF_RVV_F_OPS #undef DEF_RVV_B_OPS #undef DEF_RVV_WEXTI_OPS diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..f429f12dc18 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -246,6 +246,63 @@ static const rvv_type_info iu_ops[] = { #include "riscv-vector-builtins-types.def" {NUM_VECTOR_TYPES, 0}}; +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i8_ops[] = { +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i16_ops[] = { +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i32_ops[] = { +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u8_ops[] = { +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u16_ops[] = { +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u32_ops[] = { +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu8_ops[] = { +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu16_ops[] = { +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu32_ops[] = { +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + /* A list of all types will be registered for intrinsic functions. */ static const rvv_type_info all_ops[] = { #define DEF_RVV_I_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, @@ -934,6 +991,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1538,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2729,222 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_size_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_size_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_size_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_size_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_size_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_size_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_index_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_index_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_index_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_index_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_index_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_index_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew8_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_index_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew16_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_index_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew32_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_index_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type, + * function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type) + * function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type) + * function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_size_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_size_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_size_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +3123,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..a8ee39a3cb2 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -114,6 +114,7 @@ static const unsigned int CP_WRITE_CSR = 1U << 5; enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ @@ -234,6 +235,8 @@ struct function_group_info { case VECTOR_EXT: return TARGET_VECTOR; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; case ZVBB_EXT: return TARGET_ZVBB; case ZVBB_OR_ZVKB_EXT: diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..667820d4c3e --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (th_vlb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlbu, th_loadstore_width, full_preds, u8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlhu, th_loadstore_width, full_preds, u16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlwu, th_loadstore_width, full_preds, u32_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vsb, th_loadstore_width, none_m_preds, iu8_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vsh, th_loadstore_width, none_m_preds, iu16_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vsw, th_loadstore_width, none_m_preds, iu32_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vlsb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsbu, th_loadstore_width, full_preds, u8_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlshu, th_loadstore_width, full_preds, u16_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlswu, th_loadstore_width, full_preds, u32_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssb, th_loadstore_width, none_m_preds, iu8_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssh, th_loadstore_width, none_m_preds, iu16_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssw, th_loadstore_width, none_m_preds, iu32_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlxb, th_indexed_loadstore_width, full_preds, i8_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxh, th_indexed_loadstore_width, full_preds, i16_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxw, th_indexed_loadstore_width, full_preds, i32_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxbu, th_indexed_loadstore_width, full_preds, u8_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxhu, th_indexed_loadstore_width, full_preds, u16_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxwu, th_indexed_loadstore_width, full_preds, u32_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxb, th_indexed_loadstore_width, none_m_preds, iu8_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxh, th_indexed_loadstore_width, none_m_preds, iu16_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxw, th_indexed_loadstore_width, none_m_preds, iu32_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxb, th_indexed_loadstore_width, none_m_preds, iu8_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxh, th_indexed_loadstore_width, none_m_preds, iu16_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxw, th_indexed_loadstore_width, none_m_preds, iu32_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..0f3700d9269 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,168 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..4e192bbf025 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..1538afec68e --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..bf4924a1d76 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..8c451845175 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..0f5b09684a5 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..aaa75be023d --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} \ No newline at end of file -- 2.17.1 [-- Attachment #2: temp4cj.png --] [-- Type: application/octet-stream, Size: 500533 bytes --] [-- Attachment #3: temp4cj.png --] [-- Type: application/octet-stream, Size: 676546 bytes --] ^ permalink raw reply [flat|nested] 133+ messages in thread
* [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-04 2:34 ` [PATCH v4] " Jun Sha (Joshua) 2024-01-10 9:27 ` [PATCH v5] " Jun Sha (Joshua) @ 2024-01-10 9:31 ` Jun Sha (Joshua) 2024-01-11 8:46 ` Jun Sha (Joshua) 2024-01-11 9:52 ` [PATCH " Jun Sha (Joshua) 1 sibling, 2 replies; 133+ messages in thread From: Jun Sha (Joshua) @ 2024-01-10 9:31 UTC (permalink / raw) To: gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, Jun Sha (Joshua), Jin Ma, Xianmiao Qu This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-functions.def (vlsegff): Include thead-vector-builtins-functions.def. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins-types.def (DEF_RVV_I8_OPS): Add datatypes for XTheadVector. (DEF_RVV_I16_OPS): Add datatypes for XTheadVector. (DEF_RVV_I32_OPS): Add datatypes for XTheadVector. (DEF_RVV_U8_OPS): Add datatypes for XTheadVector. (DEF_RVV_U16_OPS): Add datatypes for XTheadVector. (DEF_RVV_U32_OPS): Add datatypes for XTheadVector. (vint8m1_t): Add datatypes for XTheadVector. (vint8m2_t): Likewise. (vint8m4_t): Likewise. (vint8m8_t): Likewise. (vint16m1_t): Likewise. (vint16m2_t): Likewise. (vint16m4_t): Likewise. (vint16m8_t): Likewise. (vint32m1_t): Likewise. (vint32m2_t): Likewise. (vint32m4_t): Likewise. (vint32m8_t): Likewise. (vint64m1_t): Likewise. (vint64m2_t): Likewise. (vint64m4_t): Likewise. (vint64m8_t): Likewise. (vuint8m1_t): Likewise. (vuint8m2_t): Likewise. (vuint8m4_t): Likewise. (vuint8m8_t): Likewise. (vuint16m1_t): Likewise. (vuint16m2_t): Likewise. (vuint16m4_t): Likewise. (vuint16m8_t): Likewise. (vuint32m1_t): Likewise. (vuint32m2_t): Likewise. (vuint32m4_t): Likewise. (vuint32m8_t): Likewise. (vuint64m1_t): Likewise. (vuint64m2_t): Likewise. (vuint64m4_t): Likewise. (vuint64m8_t): Likewise. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_I8_OPS): Add datatypes for XTheadVector. (DEF_RVV_I16_OPS): Add datatypes for XTheadVector. (DEF_RVV_I32_OPS): Add datatypes for XTheadVector. (DEF_RVV_U8_OPS): Add datatypes for XTheadVector. (DEF_RVV_U16_OPS): Add datatypes for XTheadVector. (DEF_RVV_U32_OPS): Add datatypes for XTheadVector. * config/riscv/thead-vector-builtins-functions.def: New file. * config/riscv/thead-vector.md: Add new patterns. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 ++ .../riscv/riscv-vector-builtins-shapes.cc | 98 ++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + .../riscv/riscv-vector-builtins-types.def | 120 +++++++ gcc/config/riscv/riscv-vector-builtins.cc | 311 ++++++++++++++++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 253 ++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 ++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 ++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 ++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 ++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 ++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 ++++ 16 files changed, 1406 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 46f1a1da33e..5f44f31a12b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2125,6 +2125,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2587,6 +2664,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> th_vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> th_vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> th_vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> th_vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> th_vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> th_vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> th_vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> th_vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> th_vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> th_vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> th_vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> th_vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> th_vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> th_vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> th_vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> th_vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> th_vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> th_vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> th_vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> th_vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> th_vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> th_vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> th_vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> th_vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> th_vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> th_vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> th_vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> th_vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> th_vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> th_vsuxw_obj; +static CONSTEXPR const th_extract th_vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2878,6 +2986,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (th_vlb) +BASE (th_vlh) +BASE (th_vlw) +BASE (th_vlbu) +BASE (th_vlhu) +BASE (th_vlwu) +BASE (th_vsb) +BASE (th_vsh) +BASE (th_vsw) +BASE (th_vlsb) +BASE (th_vlsh) +BASE (th_vlsw) +BASE (th_vlsbu) +BASE (th_vlshu) +BASE (th_vlswu) +BASE (th_vssb) +BASE (th_vssh) +BASE (th_vssw) +BASE (th_vlxb) +BASE (th_vlxh) +BASE (th_vlxw) +BASE (th_vlxbu) +BASE (th_vlxhu) +BASE (th_vlxwu) +BASE (th_vsxb) +BASE (th_vsxh) +BASE (th_vsxw) +BASE (th_vsuxb) +BASE (th_vsuxh) +BASE (th_vsuxw) +BASE (th_vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 1122e3801a7..df43adf9a17 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -299,6 +299,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const th_vlb; +extern const function_base *const th_vlh; +extern const function_base *const th_vlw; +extern const function_base *const th_vlbu; +extern const function_base *const th_vlhu; +extern const function_base *const th_vlwu; +extern const function_base *const th_vsb; +extern const function_base *const th_vsh; +extern const function_base *const th_vsw; +extern const function_base *const th_vlsb; +extern const function_base *const th_vlsh; +extern const function_base *const th_vlsw; +extern const function_base *const th_vlsbu; +extern const function_base *const th_vlshu; +extern const function_base *const th_vlswu; +extern const function_base *const th_vssb; +extern const function_base *const th_vssh; +extern const function_base *const th_vssw; +extern const function_base *const th_vlxb; +extern const function_base *const th_vlxh; +extern const function_base *const th_vlxw; +extern const function_base *const th_vlxbu; +extern const function_base *const th_vlxhu; +extern const function_base *const th_vlxwu; +extern const function_base *const th_vsxb; +extern const function_base *const th_vsxh; +extern const function_base *const th_vsxw; +extern const function_base *const th_vsuxb; +extern const function_base *const th_vsuxh; +extern const function_base *const th_vsuxw; +extern const function_base *const th_vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..489a95cf684 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,86 @@ struct indexed_loadstore_def : public function_shape } }; +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_base_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_one (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_base_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +712,21 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_base_name (instance.base_name); + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1189,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1203,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins-types.def b/gcc/config/riscv/riscv-vector-builtins-types.def index 61019a56844..abfeb4fcd9b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-types.def +++ b/gcc/config/riscv/riscv-vector-builtins-types.def @@ -24,12 +24,48 @@ along with GCC; see the file COPYING3. If not see #define DEF_RVV_I_OPS(TYPE, REQUIRE) #endif +/* Use "DEF_RVV_I8_OPS" macro include some signed integer (i8/i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I8_OPS +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_I16_OPS" macro include some signed integer (i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I16_OPS +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_I32_OPS" macro include some signed integer (i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I32_OPS +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) +#endif + /* Use "DEF_RVV_U_OPS" macro include all unsigned integer which will be iterated and registered as intrinsic functions. */ #ifndef DEF_RVV_U_OPS #define DEF_RVV_U_OPS(TYPE, REQUIRE) #endif +/* Use "DEF_RVV_U8_OPS" macro include some unsigned integer (i8/i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U8_OPS +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_U16_OPS" macro include some unsigned integer (i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U16_OPS +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_U32_OPS" macro include some unsigned integer (i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U32_OPS +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) +#endif + /* Use "DEF_RVV_F_OPS" macro include all floating-point which will be iterated and registered as intrinsic functions. */ #ifndef DEF_RVV_F_OPS @@ -374,6 +410,45 @@ DEF_RVV_I_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) DEF_RVV_I_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) DEF_RVV_I_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint8m1_t, 0) +DEF_RVV_I8_OPS (vint8m2_t, 0) +DEF_RVV_I8_OPS (vint8m4_t, 0) +DEF_RVV_I8_OPS (vint8m8_t, 0) +DEF_RVV_I8_OPS (vint16m1_t, 0) +DEF_RVV_I8_OPS (vint16m2_t, 0) +DEF_RVV_I8_OPS (vint16m4_t, 0) +DEF_RVV_I8_OPS (vint16m8_t, 0) +DEF_RVV_I8_OPS (vint32m1_t, 0) +DEF_RVV_I8_OPS (vint32m2_t, 0) +DEF_RVV_I8_OPS (vint32m4_t, 0) +DEF_RVV_I8_OPS (vint32m8_t, 0) +DEF_RVV_I8_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_I16_OPS (vint16m1_t, 0) +DEF_RVV_I16_OPS (vint16m2_t, 0) +DEF_RVV_I16_OPS (vint16m4_t, 0) +DEF_RVV_I16_OPS (vint16m8_t, 0) +DEF_RVV_I16_OPS (vint32m1_t, 0) +DEF_RVV_I16_OPS (vint32m2_t, 0) +DEF_RVV_I16_OPS (vint32m4_t, 0) +DEF_RVV_I16_OPS (vint32m8_t, 0) +DEF_RVV_I16_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_I32_OPS (vint32m1_t, 0) +DEF_RVV_I32_OPS (vint32m2_t, 0) +DEF_RVV_I32_OPS (vint32m4_t, 0) +DEF_RVV_I32_OPS (vint32m8_t, 0) +DEF_RVV_I32_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + DEF_RVV_U_OPS (vuint8mf8_t, RVV_REQUIRE_MIN_VLEN_64) DEF_RVV_U_OPS (vuint8mf4_t, 0) DEF_RVV_U_OPS (vuint8mf2_t, 0) @@ -397,6 +472,45 @@ DEF_RVV_U_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) DEF_RVV_U_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) DEF_RVV_U_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint8m1_t, 0) +DEF_RVV_U8_OPS (vuint8m2_t, 0) +DEF_RVV_U8_OPS (vuint8m4_t, 0) +DEF_RVV_U8_OPS (vuint8m8_t, 0) +DEF_RVV_U8_OPS (vuint16m1_t, 0) +DEF_RVV_U8_OPS (vuint16m2_t, 0) +DEF_RVV_U8_OPS (vuint16m4_t, 0) +DEF_RVV_U8_OPS (vuint16m8_t, 0) +DEF_RVV_U8_OPS (vuint32m1_t, 0) +DEF_RVV_U8_OPS (vuint32m2_t, 0) +DEF_RVV_U8_OPS (vuint32m4_t, 0) +DEF_RVV_U8_OPS (vuint32m8_t, 0) +DEF_RVV_U8_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_U16_OPS (vuint16m1_t, 0) +DEF_RVV_U16_OPS (vuint16m2_t, 0) +DEF_RVV_U16_OPS (vuint16m4_t, 0) +DEF_RVV_U16_OPS (vuint16m8_t, 0) +DEF_RVV_U16_OPS (vuint32m1_t, 0) +DEF_RVV_U16_OPS (vuint32m2_t, 0) +DEF_RVV_U16_OPS (vuint32m4_t, 0) +DEF_RVV_U16_OPS (vuint32m8_t, 0) +DEF_RVV_U16_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_U32_OPS (vuint32m1_t, 0) +DEF_RVV_U32_OPS (vuint32m2_t, 0) +DEF_RVV_U32_OPS (vuint32m4_t, 0) +DEF_RVV_U32_OPS (vuint32m8_t, 0) +DEF_RVV_U32_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + DEF_RVV_F_OPS (vfloat16mf4_t, RVV_REQUIRE_ELEN_FP_16 | RVV_REQUIRE_MIN_VLEN_64) DEF_RVV_F_OPS (vfloat16mf2_t, RVV_REQUIRE_ELEN_FP_16) DEF_RVV_F_OPS (vfloat16m1_t, RVV_REQUIRE_ELEN_FP_16) @@ -1379,7 +1493,13 @@ DEF_RVV_CRYPTO_SEW64_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) DEF_RVV_CRYPTO_SEW64_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) #undef DEF_RVV_I_OPS +#undef DEF_RVV_I8_OPS +#undef DEF_RVV_I16_OPS +#undef DEF_RVV_I32_OPS #undef DEF_RVV_U_OPS +#undef DEF_RVV_U8_OPS +#undef DEF_RVV_U16_OPS +#undef DEF_RVV_U32_OPS #undef DEF_RVV_F_OPS #undef DEF_RVV_B_OPS #undef DEF_RVV_WEXTI_OPS diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..f429f12dc18 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -246,6 +246,63 @@ static const rvv_type_info iu_ops[] = { #include "riscv-vector-builtins-types.def" {NUM_VECTOR_TYPES, 0}}; +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i8_ops[] = { +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i16_ops[] = { +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i32_ops[] = { +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u8_ops[] = { +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u16_ops[] = { +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u32_ops[] = { +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu8_ops[] = { +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu16_ops[] = { +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu32_ops[] = { +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + /* A list of all types will be registered for intrinsic functions. */ static const rvv_type_info all_ops[] = { #define DEF_RVV_I_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, @@ -934,6 +991,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1538,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2729,222 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_size_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_size_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_size_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_size_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_size_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_size_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_index_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_index_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_index_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_index_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_index_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_index_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew8_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_index_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew16_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_index_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew32_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_index_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type, + * function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type) + * function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type) + * function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_size_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_size_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_size_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +3123,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..a8ee39a3cb2 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -114,6 +114,7 @@ static const unsigned int CP_WRITE_CSR = 1U << 5; enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ @@ -234,6 +235,8 @@ struct function_group_info { case VECTOR_EXT: return TARGET_VECTOR; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; case ZVBB_EXT: return TARGET_ZVBB; case ZVBB_OR_ZVKB_EXT: diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..667820d4c3e --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (th_vlb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlbu, th_loadstore_width, full_preds, u8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlhu, th_loadstore_width, full_preds, u16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlwu, th_loadstore_width, full_preds, u32_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vsb, th_loadstore_width, none_m_preds, iu8_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vsh, th_loadstore_width, none_m_preds, iu16_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vsw, th_loadstore_width, none_m_preds, iu32_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vlsb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsbu, th_loadstore_width, full_preds, u8_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlshu, th_loadstore_width, full_preds, u16_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlswu, th_loadstore_width, full_preds, u32_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssb, th_loadstore_width, none_m_preds, iu8_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssh, th_loadstore_width, none_m_preds, iu16_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssw, th_loadstore_width, none_m_preds, iu32_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlxb, th_indexed_loadstore_width, full_preds, i8_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxh, th_indexed_loadstore_width, full_preds, i16_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxw, th_indexed_loadstore_width, full_preds, i32_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxbu, th_indexed_loadstore_width, full_preds, u8_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxhu, th_indexed_loadstore_width, full_preds, u16_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxwu, th_indexed_loadstore_width, full_preds, u32_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxb, th_indexed_loadstore_width, none_m_preds, iu8_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxh, th_indexed_loadstore_width, none_m_preds, iu16_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxw, th_indexed_loadstore_width, none_m_preds, iu32_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxb, th_indexed_loadstore_width, none_m_preds, iu8_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxh, th_indexed_loadstore_width, none_m_preds, iu16_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxw, th_indexed_loadstore_width, none_m_preds, iu32_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..0f3700d9269 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,168 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..4e192bbf025 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..6cf1513da76 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..0f59a4d8ea8 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..fbad9e8f367 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..6d2d8a59ad7 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..2d77bc6e922 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-10 9:31 ` [PATCH " Jun Sha (Joshua) @ 2024-01-11 8:46 ` Jun Sha (Joshua) 2024-01-11 9:07 ` juzhe.zhong 2024-01-11 9:52 ` [PATCH " Jun Sha (Joshua) 1 sibling, 1 reply; 133+ messages in thread From: Jun Sha (Joshua) @ 2024-01-11 8:46 UTC (permalink / raw) To: gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, Jun Sha (Joshua), Jin Ma, Xianmiao Qu This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (class th_extract): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (struct th_extract_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): * config/riscv/riscv-vector-builtins.h (enum required_ext): (struct function_group_info): * config/riscv/t-riscv: Add thead-vector-builtins-functions.def * config/riscv/thead-vector.md (@pred_mov_width<vlmem_op_attr><mode>): Add new patterns. (*pred_mov_width<vlmem_op_attr><mode>): Likewise. (@pred_store_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_load_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_store_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_load_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>): (@pred_th_extract<mode>): Likewise. (*pred_th_extract<mode>): Likewise. * config/riscv/thead-vector-builtins-functions.def: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 +++ .../riscv/riscv-vector-builtins-shapes.cc | 160 +++++++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + gcc/config/riscv/riscv-vector-builtins.cc | 70 +++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 253 ++++++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 +++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 +++++ 15 files changed, 1107 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 46f1a1da33e..3eba7943757 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2125,6 +2125,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2587,6 +2664,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> vsuxw_obj; +static CONSTEXPR const th_extract vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2878,6 +2986,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (vlb) +BASE (vlh) +BASE (vlw) +BASE (vlbu) +BASE (vlhu) +BASE (vlwu) +BASE (vsb) +BASE (vsh) +BASE (vsw) +BASE (vlsb) +BASE (vlsh) +BASE (vlsw) +BASE (vlsbu) +BASE (vlshu) +BASE (vlswu) +BASE (vssb) +BASE (vssh) +BASE (vssw) +BASE (vlxb) +BASE (vlxh) +BASE (vlxw) +BASE (vlxbu) +BASE (vlxhu) +BASE (vlxwu) +BASE (vsxb) +BASE (vsxh) +BASE (vsxw) +BASE (vsuxb) +BASE (vsuxh) +BASE (vsuxw) +BASE (vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 1122e3801a7..565a0311d2b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -299,6 +299,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const vlb; +extern const function_base *const vlh; +extern const function_base *const vlw; +extern const function_base *const vlbu; +extern const function_base *const vlhu; +extern const function_base *const vlwu; +extern const function_base *const vsb; +extern const function_base *const vsh; +extern const function_base *const vsw; +extern const function_base *const vlsb; +extern const function_base *const vlsh; +extern const function_base *const vlsw; +extern const function_base *const vlsbu; +extern const function_base *const vlshu; +extern const function_base *const vlswu; +extern const function_base *const vssb; +extern const function_base *const vssh; +extern const function_base *const vssw; +extern const function_base *const vlxb; +extern const function_base *const vlxh; +extern const function_base *const vlxw; +extern const function_base *const vlxbu; +extern const function_base *const vlxhu; +extern const function_base *const vlxwu; +extern const function_base *const vsxb; +extern const function_base *const vsxh; +extern const function_base *const vsxw; +extern const function_base *const vsuxb; +extern const function_base *const vsuxh; +extern const function_base *const vsuxw; +extern const function_base *const vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..8e90b17a94b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,146 @@ struct indexed_loadstore_def : public function_shape } }; +/* Add one function instance for GROUP, using operand suffix at index OI, + mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ +static void +build_th_loadstore (function_builder &b, const function_group_info &group, + unsigned int pred_idx, unsigned int vec_type_idx) +{ + auto_vec<tree, 5> argument_types; + function_instance function_instance (group.base_name, *group.base, + *group.shape, + group.ops_infos.types[vec_type_idx], + group.preds[pred_idx], &group.ops_infos); + tree return_type = group.ops_infos.ret.get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + b.allocate_argument_types (function_instance, argument_types); + b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + + tree type = builtin_types[group.ops_infos.types[vec_type_idx].index].vector; + if (strstr (group.base_name, "l") + && strstr (group.base_name, "u") + && !TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + if (strstr (group.base_name, "l") + && !strstr (group.base_name, "u") + && TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + machine_mode mode = TYPE_MODE (type); + int sew = GET_MODE_BITSIZE (GET_MODE_INNER (mode)); + if (strstr (group.base_name, "h") && sew == 8) + return; + + if (strstr (group.base_name, "w") && (sew == 8 || sew ==16)) + return; + + b.add_overloaded_function (function_instance, *group.shape); + b.add_unique_function (function_instance, (*group.shape), return_type, + argument_types); +} + +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +772,23 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1251,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1265,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..44b9fec1898 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -934,6 +934,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1481,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2672,38 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * index_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +2882,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..a8ee39a3cb2 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -114,6 +114,7 @@ static const unsigned int CP_WRITE_CSR = 1U << 5; enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ @@ -234,6 +235,8 @@ struct function_group_info { case VECTOR_EXT: return TARGET_VECTOR; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; case ZVBB_EXT: return TARGET_ZVBB; case ZVBB_OR_ZVKB_EXT: diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..fd3ba29bae9 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlhu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlwu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vsb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vlsb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlshu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlswu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vssb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vlxb, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxh, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxw, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxbu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxhu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxwu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vsxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..0f3700d9269 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,168 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..3c12c124597 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..30bef369375 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..3c8b5ccc16b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..b7c00404f18 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..17a53012acf --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..b187cfc852b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-11 8:46 ` Jun Sha (Joshua) @ 2024-01-11 9:07 ` juzhe.zhong 2024-01-11 9:11 ` Re：[PATCH " joshua 2024-01-11 9:54 ` Re：[PATCH " joshua 0 siblings, 2 replies; 133+ messages in thread From: juzhe.zhong @ 2024-01-11 9:07 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, cooper.joshua, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 56784 bytes --] enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ Add theadvector to the end of the enum. + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; Same. + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" You can use whole_reg_to_reg_move_p Btw, I review again : RISC-V: Handle differences between XTheadvector and Vector (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3713,7 +3744,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3734,7 +3765,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") Why do you add these !TARGERT_XTHEADVECRTOR ? juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-11 16:46 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (class th_extract): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (struct th_extract_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): * config/riscv/riscv-vector-builtins.h (enum required_ext): (struct function_group_info): * config/riscv/t-riscv: Add thead-vector-builtins-functions.def * config/riscv/thead-vector.md (@pred_mov_width<vlmem_op_attr><mode>): Add new patterns. (*pred_mov_width<vlmem_op_attr><mode>): Likewise. (@pred_store_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_load_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_store_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_load_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>): (@pred_th_extract<mode>): Likewise. (*pred_th_extract<mode>): Likewise. * config/riscv/thead-vector-builtins-functions.def: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 +++ .../riscv/riscv-vector-builtins-shapes.cc | 160 +++++++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + gcc/config/riscv/riscv-vector-builtins.cc | 70 +++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 253 ++++++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 +++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 +++++ 15 files changed, 1107 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 46f1a1da33e..3eba7943757 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2125,6 +2125,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2587,6 +2664,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> vsuxw_obj; +static CONSTEXPR const th_extract vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2878,6 +2986,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (vlb) +BASE (vlh) +BASE (vlw) +BASE (vlbu) +BASE (vlhu) +BASE (vlwu) +BASE (vsb) +BASE (vsh) +BASE (vsw) +BASE (vlsb) +BASE (vlsh) +BASE (vlsw) +BASE (vlsbu) +BASE (vlshu) +BASE (vlswu) +BASE (vssb) +BASE (vssh) +BASE (vssw) +BASE (vlxb) +BASE (vlxh) +BASE (vlxw) +BASE (vlxbu) +BASE (vlxhu) +BASE (vlxwu) +BASE (vsxb) +BASE (vsxh) +BASE (vsxw) +BASE (vsuxb) +BASE (vsuxh) +BASE (vsuxw) +BASE (vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 1122e3801a7..565a0311d2b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -299,6 +299,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const vlb; +extern const function_base *const vlh; +extern const function_base *const vlw; +extern const function_base *const vlbu; +extern const function_base *const vlhu; +extern const function_base *const vlwu; +extern const function_base *const vsb; +extern const function_base *const vsh; +extern const function_base *const vsw; +extern const function_base *const vlsb; +extern const function_base *const vlsh; +extern const function_base *const vlsw; +extern const function_base *const vlsbu; +extern const function_base *const vlshu; +extern const function_base *const vlswu; +extern const function_base *const vssb; +extern const function_base *const vssh; +extern const function_base *const vssw; +extern const function_base *const vlxb; +extern const function_base *const vlxh; +extern const function_base *const vlxw; +extern const function_base *const vlxbu; +extern const function_base *const vlxhu; +extern const function_base *const vlxwu; +extern const function_base *const vsxb; +extern const function_base *const vsxh; +extern const function_base *const vsxw; +extern const function_base *const vsuxb; +extern const function_base *const vsuxh; +extern const function_base *const vsuxw; +extern const function_base *const vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..8e90b17a94b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,146 @@ struct indexed_loadstore_def : public function_shape } }; +/* Add one function instance for GROUP, using operand suffix at index OI, + mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ +static void +build_th_loadstore (function_builder &b, const function_group_info &group, + unsigned int pred_idx, unsigned int vec_type_idx) +{ + auto_vec<tree, 5> argument_types; + function_instance function_instance (group.base_name, *group.base, + *group.shape, + group.ops_infos.types[vec_type_idx], + group.preds[pred_idx], &group.ops_infos); + tree return_type = group.ops_infos.ret.get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + b.allocate_argument_types (function_instance, argument_types); + b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + + tree type = builtin_types[group.ops_infos.types[vec_type_idx].index].vector; + if (strstr (group.base_name, "l") + && strstr (group.base_name, "u") + && !TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + if (strstr (group.base_name, "l") + && !strstr (group.base_name, "u") + && TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + machine_mode mode = TYPE_MODE (type); + int sew = GET_MODE_BITSIZE (GET_MODE_INNER (mode)); + if (strstr (group.base_name, "h") && sew == 8) + return; + + if (strstr (group.base_name, "w") && (sew == 8 || sew ==16)) + return; + + b.add_overloaded_function (function_instance, *group.shape); + b.add_unique_function (function_instance, (*group.shape), return_type, + argument_types); +} + +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +772,23 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1251,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1265,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..44b9fec1898 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -934,6 +934,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1481,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2672,38 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * index_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +2882,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..a8ee39a3cb2 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -114,6 +114,7 @@ static const unsigned int CP_WRITE_CSR = 1U << 5; enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ @@ -234,6 +235,8 @@ struct function_group_info { case VECTOR_EXT: return TARGET_VECTOR; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; case ZVBB_EXT: return TARGET_ZVBB; case ZVBB_OR_ZVKB_EXT: diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..fd3ba29bae9 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlhu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlwu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vsb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vlsb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlshu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlswu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vssb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vlxb, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxh, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxw, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxbu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxhu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxwu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vsxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..0f3700d9269 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,168 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..3c12c124597 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..30bef369375 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..3c8b5ccc16b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..b7c00404f18 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..17a53012acf --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..b187cfc852b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-11 9:07 ` juzhe.zhong @ 2024-01-11 9:11 ` joshua 2024-01-11 9:14 ` joshua 2024-01-11 9:54 ` Re：[PATCH " joshua 1 sibling, 1 reply; 133+ messages in thread From: joshua @ 2024-01-11 9:11 UTC (permalink / raw) To: juzhe.zhong, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu sext/zext will be generated in O2 even without corresponding intrinsics. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:07 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ Add theadvector to the end of the enum. + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; Same. + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" You can use whole_reg_to_reg_move_p Btw, I review again : RISC-V: Handle differences between XTheadvector and Vector (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3713,7 +3744,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3734,7 +3765,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") Why do you add these !TARGERT_XTHEADVECRTOR ? juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-11 16:46 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (class th_extract): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (struct th_extract_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): * config/riscv/riscv-vector-builtins.h (enum required_ext): (struct function_group_info): * config/riscv/t-riscv: Add thead-vector-builtins-functions.def * config/riscv/thead-vector.md (@pred_mov_width<vlmem_op_attr><mode>): Add new patterns. (*pred_mov_width<vlmem_op_attr><mode>): Likewise. (@pred_store_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_load_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_store_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_load_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>): (@pred_th_extract<mode>): Likewise. (*pred_th_extract<mode>): Likewise. * config/riscv/thead-vector-builtins-functions.def: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 +++ .../riscv/riscv-vector-builtins-shapes.cc | 160 +++++++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + gcc/config/riscv/riscv-vector-builtins.cc | 70 +++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 253 ++++++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 +++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 +++++ 15 files changed, 1107 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 46f1a1da33e..3eba7943757 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2125,6 +2125,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2587,6 +2664,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> vsuxw_obj; +static CONSTEXPR const th_extract vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2878,6 +2986,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (vlb) +BASE (vlh) +BASE (vlw) +BASE (vlbu) +BASE (vlhu) +BASE (vlwu) +BASE (vsb) +BASE (vsh) +BASE (vsw) +BASE (vlsb) +BASE (vlsh) +BASE (vlsw) +BASE (vlsbu) +BASE (vlshu) +BASE (vlswu) +BASE (vssb) +BASE (vssh) +BASE (vssw) +BASE (vlxb) +BASE (vlxh) +BASE (vlxw) +BASE (vlxbu) +BASE (vlxhu) +BASE (vlxwu) +BASE (vsxb) +BASE (vsxh) +BASE (vsxw) +BASE (vsuxb) +BASE (vsuxh) +BASE (vsuxw) +BASE (vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 1122e3801a7..565a0311d2b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -299,6 +299,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const vlb; +extern const function_base *const vlh; +extern const function_base *const vlw; +extern const function_base *const vlbu; +extern const function_base *const vlhu; +extern const function_base *const vlwu; +extern const function_base *const vsb; +extern const function_base *const vsh; +extern const function_base *const vsw; +extern const function_base *const vlsb; +extern const function_base *const vlsh; +extern const function_base *const vlsw; +extern const function_base *const vlsbu; +extern const function_base *const vlshu; +extern const function_base *const vlswu; +extern const function_base *const vssb; +extern const function_base *const vssh; +extern const function_base *const vssw; +extern const function_base *const vlxb; +extern const function_base *const vlxh; +extern const function_base *const vlxw; +extern const function_base *const vlxbu; +extern const function_base *const vlxhu; +extern const function_base *const vlxwu; +extern const function_base *const vsxb; +extern const function_base *const vsxh; +extern const function_base *const vsxw; +extern const function_base *const vsuxb; +extern const function_base *const vsuxh; +extern const function_base *const vsuxw; +extern const function_base *const vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..8e90b17a94b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,146 @@ struct indexed_loadstore_def : public function_shape } }; +/* Add one function instance for GROUP, using operand suffix at index OI, + mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ +static void +build_th_loadstore (function_builder &b, const function_group_info &group, + unsigned int pred_idx, unsigned int vec_type_idx) +{ + auto_vec<tree, 5> argument_types; + function_instance function_instance (group.base_name, *group.base, + *group.shape, + group.ops_infos.types[vec_type_idx], + group.preds[pred_idx], &group.ops_infos); + tree return_type = group.ops_infos.ret.get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + b.allocate_argument_types (function_instance, argument_types); + b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + + tree type = builtin_types[group.ops_infos.types[vec_type_idx].index].vector; + if (strstr (group.base_name, "l") + && strstr (group.base_name, "u") + && !TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + if (strstr (group.base_name, "l") + && !strstr (group.base_name, "u") + && TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + machine_mode mode = TYPE_MODE (type); + int sew = GET_MODE_BITSIZE (GET_MODE_INNER (mode)); + if (strstr (group.base_name, "h") && sew == 8) + return; + + if (strstr (group.base_name, "w") && (sew == 8 || sew ==16)) + return; + + b.add_overloaded_function (function_instance, *group.shape); + b.add_unique_function (function_instance, (*group.shape), return_type, + argument_types); +} + +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +772,23 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1251,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1265,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..44b9fec1898 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -934,6 +934,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1481,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2672,38 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * index_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +2882,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..a8ee39a3cb2 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -114,6 +114,7 @@ static const unsigned int CP_WRITE_CSR = 1U << 5; enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ @@ -234,6 +235,8 @@ struct function_group_info { case VECTOR_EXT: return TARGET_VECTOR; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; case ZVBB_EXT: return TARGET_ZVBB; case ZVBB_OR_ZVKB_EXT: diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..fd3ba29bae9 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlhu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlwu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vsb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vlsb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlshu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlswu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vssb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vlxb, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxh, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxw, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxbu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxhu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxwu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vsxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..0f3700d9269 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,168 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..3c12c124597 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..30bef369375 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..3c8b5ccc16b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..b7c00404f18 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..17a53012acf --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..b187cfc852b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-11 9:11 ` Re：[PATCH " joshua @ 2024-01-11 9:14 ` joshua 2024-01-11 9:17 ` juzhe.zhong 0 siblings, 1 reply; 133+ messages in thread From: joshua @ 2024-01-11 9:14 UTC (permalink / raw) To: juzhe.zhong, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu To be specific, in CSE pass, the initial pattern will be optimized into the sext/zext pattern. ------------------------------------------------------------------ 发件人：joshua <cooper.joshua@linux.alibaba.com> 发送时间：2024年1月11日(星期四) 17:11 收件人："juzhe.zhong@rivai.ai"<juzhe.zhong@rivai.ai>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. sext/zext will be generated in O2 even without corresponding intrinsics. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:07 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ Add theadvector to the end of the enum. + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; Same. + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" You can use whole_reg_to_reg_move_p Btw, I review again : RISC-V: Handle differences between XTheadvector and Vector (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3713,7 +3744,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3734,7 +3765,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") Why do you add these !TARGERT_XTHEADVECRTOR ? juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-11 16:46 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (class th_extract): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (struct th_extract_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): * config/riscv/riscv-vector-builtins.h (enum required_ext): (struct function_group_info): * config/riscv/t-riscv: Add thead-vector-builtins-functions.def * config/riscv/thead-vector.md (@pred_mov_width<vlmem_op_attr><mode>): Add new patterns. (*pred_mov_width<vlmem_op_attr><mode>): Likewise. (@pred_store_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_load_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_store_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_load_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>): (@pred_th_extract<mode>): Likewise. (*pred_th_extract<mode>): Likewise. * config/riscv/thead-vector-builtins-functions.def: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 +++ .../riscv/riscv-vector-builtins-shapes.cc | 160 +++++++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + gcc/config/riscv/riscv-vector-builtins.cc | 70 +++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 253 ++++++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 +++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 +++++ 15 files changed, 1107 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 46f1a1da33e..3eba7943757 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2125,6 +2125,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2587,6 +2664,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> vsuxw_obj; +static CONSTEXPR const th_extract vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2878,6 +2986,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (vlb) +BASE (vlh) +BASE (vlw) +BASE (vlbu) +BASE (vlhu) +BASE (vlwu) +BASE (vsb) +BASE (vsh) +BASE (vsw) +BASE (vlsb) +BASE (vlsh) +BASE (vlsw) +BASE (vlsbu) +BASE (vlshu) +BASE (vlswu) +BASE (vssb) +BASE (vssh) +BASE (vssw) +BASE (vlxb) +BASE (vlxh) +BASE (vlxw) +BASE (vlxbu) +BASE (vlxhu) +BASE (vlxwu) +BASE (vsxb) +BASE (vsxh) +BASE (vsxw) +BASE (vsuxb) +BASE (vsuxh) +BASE (vsuxw) +BASE (vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 1122e3801a7..565a0311d2b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -299,6 +299,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const vlb; +extern const function_base *const vlh; +extern const function_base *const vlw; +extern const function_base *const vlbu; +extern const function_base *const vlhu; +extern const function_base *const vlwu; +extern const function_base *const vsb; +extern const function_base *const vsh; +extern const function_base *const vsw; +extern const function_base *const vlsb; +extern const function_base *const vlsh; +extern const function_base *const vlsw; +extern const function_base *const vlsbu; +extern const function_base *const vlshu; +extern const function_base *const vlswu; +extern const function_base *const vssb; +extern const function_base *const vssh; +extern const function_base *const vssw; +extern const function_base *const vlxb; +extern const function_base *const vlxh; +extern const function_base *const vlxw; +extern const function_base *const vlxbu; +extern const function_base *const vlxhu; +extern const function_base *const vlxwu; +extern const function_base *const vsxb; +extern const function_base *const vsxh; +extern const function_base *const vsxw; +extern const function_base *const vsuxb; +extern const function_base *const vsuxh; +extern const function_base *const vsuxw; +extern const function_base *const vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..8e90b17a94b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,146 @@ struct indexed_loadstore_def : public function_shape } }; +/* Add one function instance for GROUP, using operand suffix at index OI, + mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ +static void +build_th_loadstore (function_builder &b, const function_group_info &group, + unsigned int pred_idx, unsigned int vec_type_idx) +{ + auto_vec<tree, 5> argument_types; + function_instance function_instance (group.base_name, *group.base, + *group.shape, + group.ops_infos.types[vec_type_idx], + group.preds[pred_idx], &group.ops_infos); + tree return_type = group.ops_infos.ret.get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + b.allocate_argument_types (function_instance, argument_types); + b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + + tree type = builtin_types[group.ops_infos.types[vec_type_idx].index].vector; + if (strstr (group.base_name, "l") + && strstr (group.base_name, "u") + && !TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + if (strstr (group.base_name, "l") + && !strstr (group.base_name, "u") + && TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + machine_mode mode = TYPE_MODE (type); + int sew = GET_MODE_BITSIZE (GET_MODE_INNER (mode)); + if (strstr (group.base_name, "h") && sew == 8) + return; + + if (strstr (group.base_name, "w") && (sew == 8 || sew ==16)) + return; + + b.add_overloaded_function (function_instance, *group.shape); + b.add_unique_function (function_instance, (*group.shape), return_type, + argument_types); +} + +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +772,23 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1251,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1265,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..44b9fec1898 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -934,6 +934,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1481,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2672,38 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * index_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +2882,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..a8ee39a3cb2 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -114,6 +114,7 @@ static const unsigned int CP_WRITE_CSR = 1U << 5; enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ @@ -234,6 +235,8 @@ struct function_group_info { case VECTOR_EXT: return TARGET_VECTOR; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; case ZVBB_EXT: return TARGET_ZVBB; case ZVBB_OR_ZVKB_EXT: diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..fd3ba29bae9 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlhu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlwu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vsb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vlsb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlshu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlswu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vssb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vlxb, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxh, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxw, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxbu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxhu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxwu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vsxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..0f3700d9269 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,168 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..3c12c124597 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..30bef369375 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..3c8b5ccc16b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..b7c00404f18 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..17a53012acf --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..b187cfc852b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-11 9:14 ` joshua @ 2024-01-11 9:17 ` juzhe.zhong 2024-01-11 9:21 ` Re：Re：[PATCH " joshua 2024-01-11 9:26 ` joshua 0 siblings, 2 replies; 133+ messages in thread From: juzhe.zhong @ 2024-01-11 9:17 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 58899 bytes --] You mean which pattern optimized sext/vzext pattern? I didn't see theadvector-specific extension patterns. Could you show me? juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:14 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. To be specific, in CSE pass, the initial pattern will be optimized into the sext/zext pattern. ------------------------------------------------------------------ 发件人：joshua <cooper.joshua@linux.alibaba.com> 发送时间：2024年1月11日(星期四) 17:11 收件人："juzhe.zhong@rivai.ai"<juzhe.zhong@rivai.ai>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. sext/zext will be generated in O2 even without corresponding intrinsics. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:07 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ Add theadvector to the end of the enum. + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; Same. + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" You can use whole_reg_to_reg_move_p Btw, I review again : RISC-V: Handle differences between XTheadvector and Vector (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3713,7 +3744,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3734,7 +3765,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") Why do you add these !TARGERT_XTHEADVECRTOR ? juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-11 16:46 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (class th_extract): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (struct th_extract_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): * config/riscv/riscv-vector-builtins.h (enum required_ext): (struct function_group_info): * config/riscv/t-riscv: Add thead-vector-builtins-functions.def * config/riscv/thead-vector.md (@pred_mov_width<vlmem_op_attr><mode>): Add new patterns. (*pred_mov_width<vlmem_op_attr><mode>): Likewise. (@pred_store_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_load_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_store_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_load_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>): (@pred_th_extract<mode>): Likewise. (*pred_th_extract<mode>): Likewise. * config/riscv/thead-vector-builtins-functions.def: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 +++ .../riscv/riscv-vector-builtins-shapes.cc | 160 +++++++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + gcc/config/riscv/riscv-vector-builtins.cc | 70 +++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 253 ++++++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 +++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 +++++ 15 files changed, 1107 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 46f1a1da33e..3eba7943757 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2125,6 +2125,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2587,6 +2664,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> vsuxw_obj; +static CONSTEXPR const th_extract vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2878,6 +2986,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (vlb) +BASE (vlh) +BASE (vlw) +BASE (vlbu) +BASE (vlhu) +BASE (vlwu) +BASE (vsb) +BASE (vsh) +BASE (vsw) +BASE (vlsb) +BASE (vlsh) +BASE (vlsw) +BASE (vlsbu) +BASE (vlshu) +BASE (vlswu) +BASE (vssb) +BASE (vssh) +BASE (vssw) +BASE (vlxb) +BASE (vlxh) +BASE (vlxw) +BASE (vlxbu) +BASE (vlxhu) +BASE (vlxwu) +BASE (vsxb) +BASE (vsxh) +BASE (vsxw) +BASE (vsuxb) +BASE (vsuxh) +BASE (vsuxw) +BASE (vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 1122e3801a7..565a0311d2b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -299,6 +299,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const vlb; +extern const function_base *const vlh; +extern const function_base *const vlw; +extern const function_base *const vlbu; +extern const function_base *const vlhu; +extern const function_base *const vlwu; +extern const function_base *const vsb; +extern const function_base *const vsh; +extern const function_base *const vsw; +extern const function_base *const vlsb; +extern const function_base *const vlsh; +extern const function_base *const vlsw; +extern const function_base *const vlsbu; +extern const function_base *const vlshu; +extern const function_base *const vlswu; +extern const function_base *const vssb; +extern const function_base *const vssh; +extern const function_base *const vssw; +extern const function_base *const vlxb; +extern const function_base *const vlxh; +extern const function_base *const vlxw; +extern const function_base *const vlxbu; +extern const function_base *const vlxhu; +extern const function_base *const vlxwu; +extern const function_base *const vsxb; +extern const function_base *const vsxh; +extern const function_base *const vsxw; +extern const function_base *const vsuxb; +extern const function_base *const vsuxh; +extern const function_base *const vsuxw; +extern const function_base *const vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..8e90b17a94b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,146 @@ struct indexed_loadstore_def : public function_shape } }; +/* Add one function instance for GROUP, using operand suffix at index OI, + mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ +static void +build_th_loadstore (function_builder &b, const function_group_info &group, + unsigned int pred_idx, unsigned int vec_type_idx) +{ + auto_vec<tree, 5> argument_types; + function_instance function_instance (group.base_name, *group.base, + *group.shape, + group.ops_infos.types[vec_type_idx], + group.preds[pred_idx], &group.ops_infos); + tree return_type = group.ops_infos.ret.get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + b.allocate_argument_types (function_instance, argument_types); + b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + + tree type = builtin_types[group.ops_infos.types[vec_type_idx].index].vector; + if (strstr (group.base_name, "l") + && strstr (group.base_name, "u") + && !TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + if (strstr (group.base_name, "l") + && !strstr (group.base_name, "u") + && TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + machine_mode mode = TYPE_MODE (type); + int sew = GET_MODE_BITSIZE (GET_MODE_INNER (mode)); + if (strstr (group.base_name, "h") && sew == 8) + return; + + if (strstr (group.base_name, "w") && (sew == 8 || sew ==16)) + return; + + b.add_overloaded_function (function_instance, *group.shape); + b.add_unique_function (function_instance, (*group.shape), return_type, + argument_types); +} + +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +772,23 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1251,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1265,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..44b9fec1898 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -934,6 +934,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1481,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2672,38 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * index_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +2882,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..a8ee39a3cb2 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -114,6 +114,7 @@ static const unsigned int CP_WRITE_CSR = 1U << 5; enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ @@ -234,6 +235,8 @@ struct function_group_info { case VECTOR_EXT: return TARGET_VECTOR; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; case ZVBB_EXT: return TARGET_ZVBB; case ZVBB_OR_ZVKB_EXT: diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..fd3ba29bae9 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlhu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlwu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vsb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vlsb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlshu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlswu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vssb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vlxb, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxh, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxw, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxbu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxhu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxwu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vsxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..0f3700d9269 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,168 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..3c12c124597 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..30bef369375 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..3c8b5ccc16b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..b7c00404f18 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..17a53012acf --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..b187cfc852b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-11 9:17 ` juzhe.zhong @ 2024-01-11 9:21 ` joshua 2024-01-11 9:24 ` Re：[PATCH " juzhe.zhong 2024-01-11 9:26 ` joshua 1 sibling, 1 reply; 133+ messages in thread From: joshua @ 2024-01-11 9:21 UTC (permalink / raw) To: juzhe.zhong, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu "I didn't see theadvector-specific extension patterns. Could you show me?" They are all in the file thead-vector.md. For the sext/zext issue, perhaps I need some time to reproduce that optimization, but I can clearly remember it is related to vwmul. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:17 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. You mean which pattern optimized sext/vzext pattern? I didn't see theadvector-specific extension patterns. Could you show me? juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:14 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. To be specific, in CSE pass, the initial pattern will be optimized into the sext/zext pattern. ------------------------------------------------------------------ 发件人：joshua <cooper.joshua@linux.alibaba.com> 发送时间：2024年1月11日(星期四) 17:11 收件人："juzhe.zhong@rivai.ai"<juzhe.zhong@rivai.ai>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. sext/zext will be generated in O2 even without corresponding intrinsics. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:07 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ Add theadvector to the end of the enum. + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; Same. + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" You can use whole_reg_to_reg_move_p Btw, I review again : RISC-V: Handle differences between XTheadvector and Vector (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3713,7 +3744,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3734,7 +3765,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") Why do you add these !TARGERT_XTHEADVECRTOR ? juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-11 16:46 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (class th_extract): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (struct th_extract_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): * config/riscv/riscv-vector-builtins.h (enum required_ext): (struct function_group_info): * config/riscv/t-riscv: Add thead-vector-builtins-functions.def * config/riscv/thead-vector.md (@pred_mov_width<vlmem_op_attr><mode>): Add new patterns. (*pred_mov_width<vlmem_op_attr><mode>): Likewise. (@pred_store_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_load_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_store_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_load_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>): (@pred_th_extract<mode>): Likewise. (*pred_th_extract<mode>): Likewise. * config/riscv/thead-vector-builtins-functions.def: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 +++ .../riscv/riscv-vector-builtins-shapes.cc | 160 +++++++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + gcc/config/riscv/riscv-vector-builtins.cc | 70 +++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 253 ++++++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 +++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 +++++ 15 files changed, 1107 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 46f1a1da33e..3eba7943757 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2125,6 +2125,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2587,6 +2664,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> vsuxw_obj; +static CONSTEXPR const th_extract vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2878,6 +2986,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (vlb) +BASE (vlh) +BASE (vlw) +BASE (vlbu) +BASE (vlhu) +BASE (vlwu) +BASE (vsb) +BASE (vsh) +BASE (vsw) +BASE (vlsb) +BASE (vlsh) +BASE (vlsw) +BASE (vlsbu) +BASE (vlshu) +BASE (vlswu) +BASE (vssb) +BASE (vssh) +BASE (vssw) +BASE (vlxb) +BASE (vlxh) +BASE (vlxw) +BASE (vlxbu) +BASE (vlxhu) +BASE (vlxwu) +BASE (vsxb) +BASE (vsxh) +BASE (vsxw) +BASE (vsuxb) +BASE (vsuxh) +BASE (vsuxw) +BASE (vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 1122e3801a7..565a0311d2b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -299,6 +299,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const vlb; +extern const function_base *const vlh; +extern const function_base *const vlw; +extern const function_base *const vlbu; +extern const function_base *const vlhu; +extern const function_base *const vlwu; +extern const function_base *const vsb; +extern const function_base *const vsh; +extern const function_base *const vsw; +extern const function_base *const vlsb; +extern const function_base *const vlsh; +extern const function_base *const vlsw; +extern const function_base *const vlsbu; +extern const function_base *const vlshu; +extern const function_base *const vlswu; +extern const function_base *const vssb; +extern const function_base *const vssh; +extern const function_base *const vssw; +extern const function_base *const vlxb; +extern const function_base *const vlxh; +extern const function_base *const vlxw; +extern const function_base *const vlxbu; +extern const function_base *const vlxhu; +extern const function_base *const vlxwu; +extern const function_base *const vsxb; +extern const function_base *const vsxh; +extern const function_base *const vsxw; +extern const function_base *const vsuxb; +extern const function_base *const vsuxh; +extern const function_base *const vsuxw; +extern const function_base *const vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..8e90b17a94b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,146 @@ struct indexed_loadstore_def : public function_shape } }; +/* Add one function instance for GROUP, using operand suffix at index OI, + mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ +static void +build_th_loadstore (function_builder &b, const function_group_info &group, + unsigned int pred_idx, unsigned int vec_type_idx) +{ + auto_vec<tree, 5> argument_types; + function_instance function_instance (group.base_name, *group.base, + *group.shape, + group.ops_infos.types[vec_type_idx], + group.preds[pred_idx], &group.ops_infos); + tree return_type = group.ops_infos.ret.get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + b.allocate_argument_types (function_instance, argument_types); + b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + + tree type = builtin_types[group.ops_infos.types[vec_type_idx].index].vector; + if (strstr (group.base_name, "l") + && strstr (group.base_name, "u") + && !TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + if (strstr (group.base_name, "l") + && !strstr (group.base_name, "u") + && TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + machine_mode mode = TYPE_MODE (type); + int sew = GET_MODE_BITSIZE (GET_MODE_INNER (mode)); + if (strstr (group.base_name, "h") && sew == 8) + return; + + if (strstr (group.base_name, "w") && (sew == 8 || sew ==16)) + return; + + b.add_overloaded_function (function_instance, *group.shape); + b.add_unique_function (function_instance, (*group.shape), return_type, + argument_types); +} + +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +772,23 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1251,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1265,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..44b9fec1898 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -934,6 +934,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1481,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2672,38 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * index_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +2882,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..a8ee39a3cb2 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -114,6 +114,7 @@ static const unsigned int CP_WRITE_CSR = 1U << 5; enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ @@ -234,6 +235,8 @@ struct function_group_info { case VECTOR_EXT: return TARGET_VECTOR; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; case ZVBB_EXT: return TARGET_ZVBB; case ZVBB_OR_ZVKB_EXT: diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..fd3ba29bae9 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlhu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlwu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vsb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vlsb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlshu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlswu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vssb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vlxb, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxh, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxw, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxbu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxhu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxwu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vsxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..0f3700d9269 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,168 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..3c12c124597 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..30bef369375 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..3c8b5ccc16b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..b7c00404f18 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..17a53012acf --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..b187cfc852b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-11 9:21 ` Re：Re：[PATCH " joshua @ 2024-01-11 9:24 ` juzhe.zhong 2024-01-11 9:29 ` Re：Re：[PATCH " joshua 0 siblings, 1 reply; 133+ messages in thread From: juzhe.zhong @ 2024-01-11 9:24 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 67440 bytes --] I don't see any patterns are possible CSE into sext/zext patterns: +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:21 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. "I didn't see theadvector-specific extension patterns. Could you show me?" They are all in the file thead-vector.md. For the sext/zext issue, perhaps I need some time to reproduce that optimization, but I can clearly remember it is related to vwmul. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:17 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. You mean which pattern optimized sext/vzext pattern? I didn't see theadvector-specific extension patterns. Could you show me? juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:14 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. To be specific, in CSE pass, the initial pattern will be optimized into the sext/zext pattern. ------------------------------------------------------------------ 发件人：joshua <cooper.joshua@linux.alibaba.com> 发送时间：2024年1月11日(星期四) 17:11 收件人："juzhe.zhong@rivai.ai"<juzhe.zhong@rivai.ai>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. sext/zext will be generated in O2 even without corresponding intrinsics. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:07 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ Add theadvector to the end of the enum. + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; Same. + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" You can use whole_reg_to_reg_move_p Btw, I review again : RISC-V: Handle differences between XTheadvector and Vector (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3713,7 +3744,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3734,7 +3765,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") Why do you add these !TARGERT_XTHEADVECRTOR ? juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-11 16:46 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (class th_extract): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (struct th_extract_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): * config/riscv/riscv-vector-builtins.h (enum required_ext): (struct function_group_info): * config/riscv/t-riscv: Add thead-vector-builtins-functions.def * config/riscv/thead-vector.md (@pred_mov_width<vlmem_op_attr><mode>): Add new patterns. (*pred_mov_width<vlmem_op_attr><mode>): Likewise. (@pred_store_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_load_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_store_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_load_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>): (@pred_th_extract<mode>): Likewise. (*pred_th_extract<mode>): Likewise. * config/riscv/thead-vector-builtins-functions.def: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 +++ .../riscv/riscv-vector-builtins-shapes.cc | 160 +++++++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + gcc/config/riscv/riscv-vector-builtins.cc | 70 +++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 253 ++++++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 +++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 +++++ 15 files changed, 1107 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 46f1a1da33e..3eba7943757 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2125,6 +2125,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2587,6 +2664,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> vsuxw_obj; +static CONSTEXPR const th_extract vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2878,6 +2986,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (vlb) +BASE (vlh) +BASE (vlw) +BASE (vlbu) +BASE (vlhu) +BASE (vlwu) +BASE (vsb) +BASE (vsh) +BASE (vsw) +BASE (vlsb) +BASE (vlsh) +BASE (vlsw) +BASE (vlsbu) +BASE (vlshu) +BASE (vlswu) +BASE (vssb) +BASE (vssh) +BASE (vssw) +BASE (vlxb) +BASE (vlxh) +BASE (vlxw) +BASE (vlxbu) +BASE (vlxhu) +BASE (vlxwu) +BASE (vsxb) +BASE (vsxh) +BASE (vsxw) +BASE (vsuxb) +BASE (vsuxh) +BASE (vsuxw) +BASE (vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 1122e3801a7..565a0311d2b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -299,6 +299,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const vlb; +extern const function_base *const vlh; +extern const function_base *const vlw; +extern const function_base *const vlbu; +extern const function_base *const vlhu; +extern const function_base *const vlwu; +extern const function_base *const vsb; +extern const function_base *const vsh; +extern const function_base *const vsw; +extern const function_base *const vlsb; +extern const function_base *const vlsh; +extern const function_base *const vlsw; +extern const function_base *const vlsbu; +extern const function_base *const vlshu; +extern const function_base *const vlswu; +extern const function_base *const vssb; +extern const function_base *const vssh; +extern const function_base *const vssw; +extern const function_base *const vlxb; +extern const function_base *const vlxh; +extern const function_base *const vlxw; +extern const function_base *const vlxbu; +extern const function_base *const vlxhu; +extern const function_base *const vlxwu; +extern const function_base *const vsxb; +extern const function_base *const vsxh; +extern const function_base *const vsxw; +extern const function_base *const vsuxb; +extern const function_base *const vsuxh; +extern const function_base *const vsuxw; +extern const function_base *const vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..8e90b17a94b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,146 @@ struct indexed_loadstore_def : public function_shape } }; +/* Add one function instance for GROUP, using operand suffix at index OI, + mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ +static void +build_th_loadstore (function_builder &b, const function_group_info &group, + unsigned int pred_idx, unsigned int vec_type_idx) +{ + auto_vec<tree, 5> argument_types; + function_instance function_instance (group.base_name, *group.base, + *group.shape, + group.ops_infos.types[vec_type_idx], + group.preds[pred_idx], &group.ops_infos); + tree return_type = group.ops_infos.ret.get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + b.allocate_argument_types (function_instance, argument_types); + b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + + tree type = builtin_types[group.ops_infos.types[vec_type_idx].index].vector; + if (strstr (group.base_name, "l") + && strstr (group.base_name, "u") + && !TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + if (strstr (group.base_name, "l") + && !strstr (group.base_name, "u") + && TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + machine_mode mode = TYPE_MODE (type); + int sew = GET_MODE_BITSIZE (GET_MODE_INNER (mode)); + if (strstr (group.base_name, "h") && sew == 8) + return; + + if (strstr (group.base_name, "w") && (sew == 8 || sew ==16)) + return; + + b.add_overloaded_function (function_instance, *group.shape); + b.add_unique_function (function_instance, (*group.shape), return_type, + argument_types); +} + +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +772,23 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1251,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1265,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..44b9fec1898 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -934,6 +934,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1481,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2672,38 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * index_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +2882,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..a8ee39a3cb2 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -114,6 +114,7 @@ static const unsigned int CP_WRITE_CSR = 1U << 5; enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ @@ -234,6 +235,8 @@ struct function_group_info { case VECTOR_EXT: return TARGET_VECTOR; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; case ZVBB_EXT: return TARGET_ZVBB; case ZVBB_OR_ZVKB_EXT: diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..fd3ba29bae9 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlhu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlwu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vsb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vlsb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlshu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlswu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vssb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vlxb, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxh, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxw, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxbu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxhu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxwu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vsxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..0f3700d9269 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,168 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..3c12c124597 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..30bef369375 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..3c8b5ccc16b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..b7c00404f18 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..17a53012acf --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..b187cfc852b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-11 9:24 ` Re：[PATCH " juzhe.zhong @ 2024-01-11 9:29 ` joshua 2024-01-11 9:32 ` Re：[PATCH " juzhe.zhong 0 siblings, 1 reply; 133+ messages in thread From: joshua @ 2024-01-11 9:29 UTC (permalink / raw) To: juzhe.zhong, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu The sext/zext issue is not related to xtheadvector-special patterns. I added !TARGET_XTHEADVECTOR to sext/zext patterns in "RISC-V: Handle differences between XTheadvector and Vector" That is caused by the vwmul pattern, but I cannot reproduce it right now. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:24 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. I don't see any patterns are possible CSE into sext/zext patterns: +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:21 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. "I didn't see theadvector-specific extension patterns. Could you show me?" They are all in the file thead-vector.md. For the sext/zext issue, perhaps I need some time to reproduce that optimization, but I can clearly remember it is related to vwmul. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:17 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. You mean which pattern optimized sext/vzext pattern? I didn't see theadvector-specific extension patterns. Could you show me? juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:14 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. To be specific, in CSE pass, the initial pattern will be optimized into the sext/zext pattern. ------------------------------------------------------------------ 发件人：joshua <cooper.joshua@linux.alibaba.com> 发送时间：2024年1月11日(星期四) 17:11 收件人："juzhe.zhong@rivai.ai"<juzhe.zhong@rivai.ai>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. sext/zext will be generated in O2 even without corresponding intrinsics. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:07 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ Add theadvector to the end of the enum. + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; Same. + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" You can use whole_reg_to_reg_move_p Btw, I review again : RISC-V: Handle differences between XTheadvector and Vector (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3713,7 +3744,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3734,7 +3765,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") Why do you add these !TARGERT_XTHEADVECRTOR ? juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-11 16:46 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (class th_extract): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (struct th_extract_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): * config/riscv/riscv-vector-builtins.h (enum required_ext): (struct function_group_info): * config/riscv/t-riscv: Add thead-vector-builtins-functions.def * config/riscv/thead-vector.md (@pred_mov_width<vlmem_op_attr><mode>): Add new patterns. (*pred_mov_width<vlmem_op_attr><mode>): Likewise. (@pred_store_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_load_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_store_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_load_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>): (@pred_th_extract<mode>): Likewise. (*pred_th_extract<mode>): Likewise. * config/riscv/thead-vector-builtins-functions.def: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 +++ .../riscv/riscv-vector-builtins-shapes.cc | 160 +++++++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + gcc/config/riscv/riscv-vector-builtins.cc | 70 +++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 253 ++++++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 +++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 +++++ 15 files changed, 1107 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 46f1a1da33e..3eba7943757 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2125,6 +2125,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2587,6 +2664,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> vsuxw_obj; +static CONSTEXPR const th_extract vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2878,6 +2986,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (vlb) +BASE (vlh) +BASE (vlw) +BASE (vlbu) +BASE (vlhu) +BASE (vlwu) +BASE (vsb) +BASE (vsh) +BASE (vsw) +BASE (vlsb) +BASE (vlsh) +BASE (vlsw) +BASE (vlsbu) +BASE (vlshu) +BASE (vlswu) +BASE (vssb) +BASE (vssh) +BASE (vssw) +BASE (vlxb) +BASE (vlxh) +BASE (vlxw) +BASE (vlxbu) +BASE (vlxhu) +BASE (vlxwu) +BASE (vsxb) +BASE (vsxh) +BASE (vsxw) +BASE (vsuxb) +BASE (vsuxh) +BASE (vsuxw) +BASE (vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 1122e3801a7..565a0311d2b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -299,6 +299,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const vlb; +extern const function_base *const vlh; +extern const function_base *const vlw; +extern const function_base *const vlbu; +extern const function_base *const vlhu; +extern const function_base *const vlwu; +extern const function_base *const vsb; +extern const function_base *const vsh; +extern const function_base *const vsw; +extern const function_base *const vlsb; +extern const function_base *const vlsh; +extern const function_base *const vlsw; +extern const function_base *const vlsbu; +extern const function_base *const vlshu; +extern const function_base *const vlswu; +extern const function_base *const vssb; +extern const function_base *const vssh; +extern const function_base *const vssw; +extern const function_base *const vlxb; +extern const function_base *const vlxh; +extern const function_base *const vlxw; +extern const function_base *const vlxbu; +extern const function_base *const vlxhu; +extern const function_base *const vlxwu; +extern const function_base *const vsxb; +extern const function_base *const vsxh; +extern const function_base *const vsxw; +extern const function_base *const vsuxb; +extern const function_base *const vsuxh; +extern const function_base *const vsuxw; +extern const function_base *const vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..8e90b17a94b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,146 @@ struct indexed_loadstore_def : public function_shape } }; +/* Add one function instance for GROUP, using operand suffix at index OI, + mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ +static void +build_th_loadstore (function_builder &b, const function_group_info &group, + unsigned int pred_idx, unsigned int vec_type_idx) +{ + auto_vec<tree, 5> argument_types; + function_instance function_instance (group.base_name, *group.base, + *group.shape, + group.ops_infos.types[vec_type_idx], + group.preds[pred_idx], &group.ops_infos); + tree return_type = group.ops_infos.ret.get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + b.allocate_argument_types (function_instance, argument_types); + b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + + tree type = builtin_types[group.ops_infos.types[vec_type_idx].index].vector; + if (strstr (group.base_name, "l") + && strstr (group.base_name, "u") + && !TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + if (strstr (group.base_name, "l") + && !strstr (group.base_name, "u") + && TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + machine_mode mode = TYPE_MODE (type); + int sew = GET_MODE_BITSIZE (GET_MODE_INNER (mode)); + if (strstr (group.base_name, "h") && sew == 8) + return; + + if (strstr (group.base_name, "w") && (sew == 8 || sew ==16)) + return; + + b.add_overloaded_function (function_instance, *group.shape); + b.add_unique_function (function_instance, (*group.shape), return_type, + argument_types); +} + +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +772,23 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1251,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1265,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..44b9fec1898 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -934,6 +934,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1481,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2672,38 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * index_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +2882,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..a8ee39a3cb2 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -114,6 +114,7 @@ static const unsigned int CP_WRITE_CSR = 1U << 5; enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ @@ -234,6 +235,8 @@ struct function_group_info { case VECTOR_EXT: return TARGET_VECTOR; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; case ZVBB_EXT: return TARGET_ZVBB; case ZVBB_OR_ZVKB_EXT: diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..fd3ba29bae9 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlhu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlwu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vsb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vlsb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlshu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlswu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vssb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vlxb, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxh, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxw, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxbu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxhu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxwu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vsxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..0f3700d9269 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,168 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..3c12c124597 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..30bef369375 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..3c8b5ccc16b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..b7c00404f18 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..17a53012acf --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..b187cfc852b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-11 9:29 ` Re：Re：[PATCH " joshua @ 2024-01-11 9:32 ` juzhe.zhong 2024-01-11 9:38 ` Re：Re：[PATCH " joshua 2024-01-11 12:05 ` joshua 0 siblings, 2 replies; 133+ messages in thread From: juzhe.zhong @ 2024-01-11 9:32 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 68878 bytes --] Ok. Let's hold on "RISC-V: Handle differences between XTheadvector and Vector" patch until you can reproduce the issue for me. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:29 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. The sext/zext issue is not related to xtheadvector-special patterns. I added !TARGET_XTHEADVECTOR to sext/zext patterns in "RISC-V: Handle differences between XTheadvector and Vector" That is caused by the vwmul pattern, but I cannot reproduce it right now. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:24 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. I don't see any patterns are possible CSE into sext/zext patterns: +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:21 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. "I didn't see theadvector-specific extension patterns. Could you show me?" They are all in the file thead-vector.md. For the sext/zext issue, perhaps I need some time to reproduce that optimization, but I can clearly remember it is related to vwmul. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:17 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. You mean which pattern optimized sext/vzext pattern? I didn't see theadvector-specific extension patterns. Could you show me? juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:14 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. To be specific, in CSE pass, the initial pattern will be optimized into the sext/zext pattern. ------------------------------------------------------------------ 发件人：joshua <cooper.joshua@linux.alibaba.com> 发送时间：2024年1月11日(星期四) 17:11 收件人："juzhe.zhong@rivai.ai"<juzhe.zhong@rivai.ai>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. sext/zext will be generated in O2 even without corresponding intrinsics. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:07 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ Add theadvector to the end of the enum. + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; Same. + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" You can use whole_reg_to_reg_move_p Btw, I review again : RISC-V: Handle differences between XTheadvector and Vector (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3713,7 +3744,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3734,7 +3765,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") Why do you add these !TARGERT_XTHEADVECRTOR ? juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-11 16:46 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (class th_extract): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (struct th_extract_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): * config/riscv/riscv-vector-builtins.h (enum required_ext): (struct function_group_info): * config/riscv/t-riscv: Add thead-vector-builtins-functions.def * config/riscv/thead-vector.md (@pred_mov_width<vlmem_op_attr><mode>): Add new patterns. (*pred_mov_width<vlmem_op_attr><mode>): Likewise. (@pred_store_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_load_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_store_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_load_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>): (@pred_th_extract<mode>): Likewise. (*pred_th_extract<mode>): Likewise. * config/riscv/thead-vector-builtins-functions.def: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 +++ .../riscv/riscv-vector-builtins-shapes.cc | 160 +++++++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + gcc/config/riscv/riscv-vector-builtins.cc | 70 +++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 253 ++++++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 +++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 +++++ 15 files changed, 1107 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 46f1a1da33e..3eba7943757 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2125,6 +2125,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2587,6 +2664,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> vsuxw_obj; +static CONSTEXPR const th_extract vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2878,6 +2986,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (vlb) +BASE (vlh) +BASE (vlw) +BASE (vlbu) +BASE (vlhu) +BASE (vlwu) +BASE (vsb) +BASE (vsh) +BASE (vsw) +BASE (vlsb) +BASE (vlsh) +BASE (vlsw) +BASE (vlsbu) +BASE (vlshu) +BASE (vlswu) +BASE (vssb) +BASE (vssh) +BASE (vssw) +BASE (vlxb) +BASE (vlxh) +BASE (vlxw) +BASE (vlxbu) +BASE (vlxhu) +BASE (vlxwu) +BASE (vsxb) +BASE (vsxh) +BASE (vsxw) +BASE (vsuxb) +BASE (vsuxh) +BASE (vsuxw) +BASE (vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 1122e3801a7..565a0311d2b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -299,6 +299,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const vlb; +extern const function_base *const vlh; +extern const function_base *const vlw; +extern const function_base *const vlbu; +extern const function_base *const vlhu; +extern const function_base *const vlwu; +extern const function_base *const vsb; +extern const function_base *const vsh; +extern const function_base *const vsw; +extern const function_base *const vlsb; +extern const function_base *const vlsh; +extern const function_base *const vlsw; +extern const function_base *const vlsbu; +extern const function_base *const vlshu; +extern const function_base *const vlswu; +extern const function_base *const vssb; +extern const function_base *const vssh; +extern const function_base *const vssw; +extern const function_base *const vlxb; +extern const function_base *const vlxh; +extern const function_base *const vlxw; +extern const function_base *const vlxbu; +extern const function_base *const vlxhu; +extern const function_base *const vlxwu; +extern const function_base *const vsxb; +extern const function_base *const vsxh; +extern const function_base *const vsxw; +extern const function_base *const vsuxb; +extern const function_base *const vsuxh; +extern const function_base *const vsuxw; +extern const function_base *const vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..8e90b17a94b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,146 @@ struct indexed_loadstore_def : public function_shape } }; +/* Add one function instance for GROUP, using operand suffix at index OI, + mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ +static void +build_th_loadstore (function_builder &b, const function_group_info &group, + unsigned int pred_idx, unsigned int vec_type_idx) +{ + auto_vec<tree, 5> argument_types; + function_instance function_instance (group.base_name, *group.base, + *group.shape, + group.ops_infos.types[vec_type_idx], + group.preds[pred_idx], &group.ops_infos); + tree return_type = group.ops_infos.ret.get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + b.allocate_argument_types (function_instance, argument_types); + b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + + tree type = builtin_types[group.ops_infos.types[vec_type_idx].index].vector; + if (strstr (group.base_name, "l") + && strstr (group.base_name, "u") + && !TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + if (strstr (group.base_name, "l") + && !strstr (group.base_name, "u") + && TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + machine_mode mode = TYPE_MODE (type); + int sew = GET_MODE_BITSIZE (GET_MODE_INNER (mode)); + if (strstr (group.base_name, "h") && sew == 8) + return; + + if (strstr (group.base_name, "w") && (sew == 8 || sew ==16)) + return; + + b.add_overloaded_function (function_instance, *group.shape); + b.add_unique_function (function_instance, (*group.shape), return_type, + argument_types); +} + +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +772,23 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1251,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1265,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..44b9fec1898 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -934,6 +934,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1481,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2672,38 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * index_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +2882,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..a8ee39a3cb2 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -114,6 +114,7 @@ static const unsigned int CP_WRITE_CSR = 1U << 5; enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ @@ -234,6 +235,8 @@ struct function_group_info { case VECTOR_EXT: return TARGET_VECTOR; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; case ZVBB_EXT: return TARGET_ZVBB; case ZVBB_OR_ZVKB_EXT: diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..fd3ba29bae9 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlhu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlwu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vsb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vlsb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlshu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlswu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vssb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vlxb, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxh, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxw, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxbu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxhu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxwu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vsxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..0f3700d9269 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,168 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..3c12c124597 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..30bef369375 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..3c8b5ccc16b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..b7c00404f18 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..17a53012acf --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..b187cfc852b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-11 9:32 ` Re：[PATCH " juzhe.zhong @ 2024-01-11 9:38 ` joshua 2024-01-11 12:05 ` joshua 1 sibling, 0 replies; 133+ messages in thread From: joshua @ 2024-01-11 9:38 UTC (permalink / raw) To: juzhe.zhong, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu I think removing these changes for the "RISC-V: Handle differences between XTheadvector and Vector" patch is better. This is an extra e issue that can be handled in a seperate patch. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:32 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Ok. Let's hold on "RISC-V: Handle differences between XTheadvector and Vector" patch until you can reproduce the issue for me. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:29 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. The sext/zext issue is not related to xtheadvector-special patterns. I added !TARGET_XTHEADVECTOR to sext/zext patterns in "RISC-V: Handle differences between XTheadvector and Vector" That is caused by the vwmul pattern, but I cannot reproduce it right now. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:24 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. I don't see any patterns are possible CSE into sext/zext patterns: +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:21 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. "I didn't see theadvector-specific extension patterns. Could you show me?" They are all in the file thead-vector.md. For the sext/zext issue, perhaps I need some time to reproduce that optimization, but I can clearly remember it is related to vwmul. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:17 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. You mean which pattern optimized sext/vzext pattern? I didn't see theadvector-specific extension patterns. Could you show me? juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:14 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. To be specific, in CSE pass, the initial pattern will be optimized into the sext/zext pattern. ------------------------------------------------------------------ 发件人：joshua <cooper.joshua@linux.alibaba.com> 发送时间：2024年1月11日(星期四) 17:11 收件人："juzhe.zhong@rivai.ai"<juzhe.zhong@rivai.ai>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. sext/zext will be generated in O2 even without corresponding intrinsics. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:07 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ Add theadvector to the end of the enum. + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; Same. + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" You can use whole_reg_to_reg_move_p Btw, I review again : RISC-V: Handle differences between XTheadvector and Vector (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3713,7 +3744,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3734,7 +3765,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") Why do you add these !TARGERT_XTHEADVECRTOR ? juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-11 16:46 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (class th_extract): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (struct th_extract_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): * config/riscv/riscv-vector-builtins.h (enum required_ext): (struct function_group_info): * config/riscv/t-riscv: Add thead-vector-builtins-functions.def * config/riscv/thead-vector.md (@pred_mov_width<vlmem_op_attr><mode>): Add new patterns. (*pred_mov_width<vlmem_op_attr><mode>): Likewise. (@pred_store_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_load_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_store_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_load_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>): (@pred_th_extract<mode>): Likewise. (*pred_th_extract<mode>): Likewise. * config/riscv/thead-vector-builtins-functions.def: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 +++ .../riscv/riscv-vector-builtins-shapes.cc | 160 +++++++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + gcc/config/riscv/riscv-vector-builtins.cc | 70 +++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 253 ++++++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 +++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 +++++ 15 files changed, 1107 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 46f1a1da33e..3eba7943757 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2125,6 +2125,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2587,6 +2664,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> vsuxw_obj; +static CONSTEXPR const th_extract vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2878,6 +2986,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (vlb) +BASE (vlh) +BASE (vlw) +BASE (vlbu) +BASE (vlhu) +BASE (vlwu) +BASE (vsb) +BASE (vsh) +BASE (vsw) +BASE (vlsb) +BASE (vlsh) +BASE (vlsw) +BASE (vlsbu) +BASE (vlshu) +BASE (vlswu) +BASE (vssb) +BASE (vssh) +BASE (vssw) +BASE (vlxb) +BASE (vlxh) +BASE (vlxw) +BASE (vlxbu) +BASE (vlxhu) +BASE (vlxwu) +BASE (vsxb) +BASE (vsxh) +BASE (vsxw) +BASE (vsuxb) +BASE (vsuxh) +BASE (vsuxw) +BASE (vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 1122e3801a7..565a0311d2b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -299,6 +299,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const vlb; +extern const function_base *const vlh; +extern const function_base *const vlw; +extern const function_base *const vlbu; +extern const function_base *const vlhu; +extern const function_base *const vlwu; +extern const function_base *const vsb; +extern const function_base *const vsh; +extern const function_base *const vsw; +extern const function_base *const vlsb; +extern const function_base *const vlsh; +extern const function_base *const vlsw; +extern const function_base *const vlsbu; +extern const function_base *const vlshu; +extern const function_base *const vlswu; +extern const function_base *const vssb; +extern const function_base *const vssh; +extern const function_base *const vssw; +extern const function_base *const vlxb; +extern const function_base *const vlxh; +extern const function_base *const vlxw; +extern const function_base *const vlxbu; +extern const function_base *const vlxhu; +extern const function_base *const vlxwu; +extern const function_base *const vsxb; +extern const function_base *const vsxh; +extern const function_base *const vsxw; +extern const function_base *const vsuxb; +extern const function_base *const vsuxh; +extern const function_base *const vsuxw; +extern const function_base *const vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..8e90b17a94b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,146 @@ struct indexed_loadstore_def : public function_shape } }; +/* Add one function instance for GROUP, using operand suffix at index OI, + mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ +static void +build_th_loadstore (function_builder &b, const function_group_info &group, + unsigned int pred_idx, unsigned int vec_type_idx) +{ + auto_vec<tree, 5> argument_types; + function_instance function_instance (group.base_name, *group.base, + *group.shape, + group.ops_infos.types[vec_type_idx], + group.preds[pred_idx], &group.ops_infos); + tree return_type = group.ops_infos.ret.get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + b.allocate_argument_types (function_instance, argument_types); + b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + + tree type = builtin_types[group.ops_infos.types[vec_type_idx].index].vector; + if (strstr (group.base_name, "l") + && strstr (group.base_name, "u") + && !TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + if (strstr (group.base_name, "l") + && !strstr (group.base_name, "u") + && TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + machine_mode mode = TYPE_MODE (type); + int sew = GET_MODE_BITSIZE (GET_MODE_INNER (mode)); + if (strstr (group.base_name, "h") && sew == 8) + return; + + if (strstr (group.base_name, "w") && (sew == 8 || sew ==16)) + return; + + b.add_overloaded_function (function_instance, *group.shape); + b.add_unique_function (function_instance, (*group.shape), return_type, + argument_types); +} + +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +772,23 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1251,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1265,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..44b9fec1898 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -934,6 +934,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1481,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2672,38 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * index_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +2882,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..a8ee39a3cb2 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -114,6 +114,7 @@ static const unsigned int CP_WRITE_CSR = 1U << 5; enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ @@ -234,6 +235,8 @@ struct function_group_info { case VECTOR_EXT: return TARGET_VECTOR; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; case ZVBB_EXT: return TARGET_ZVBB; case ZVBB_OR_ZVKB_EXT: diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..fd3ba29bae9 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlhu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlwu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vsb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vlsb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlshu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlswu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vssb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vlxb, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxh, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxw, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxbu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxhu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxwu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vsxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..0f3700d9269 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,168 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..3c12c124597 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..30bef369375 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..3c8b5ccc16b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..b7c00404f18 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..17a53012acf --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..b187cfc852b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-11 9:32 ` Re：[PATCH " juzhe.zhong 2024-01-11 9:38 ` Re：Re：[PATCH " joshua @ 2024-01-11 12:05 ` joshua 2024-01-11 12:13 ` Re：[PATCH " juzhe.zhong 1 sibling, 1 reply; 133+ messages in thread From: joshua @ 2024-01-11 12:05 UTC (permalink / raw) To: juzhe.zhong, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu without CSE: (insn 16 15 17 2 (set (reg/v:RVVM2HI 137 [ output_var_0 ]) (if_then_else:RVVM2HI (unspec:RVVMF8BI [ (const_vector:RVVMF8BI repeat [ (const_int 1 [0x1]) ]) (reg:DI 146) (const_int 2 [0x2]) repeated x2 (const_int 0 [0]) (reg:SI 66 vl) (reg:SI 67 vtype) ] UNSPEC_VPREDICATE) (mult:RVVM2HI (sign_extend:RVVM2HI (reg/v:RVVM1QI 136 [ op1 ])) (sign_extend:RVVM2HI (vec_duplicate:RVVM1QI (reg:QI 147)))) (unspec:RVVM2HI [ (reg:SI 0 zero) ] UNSPEC_VUNDEF))) {pred_dual_widen_mulsrvvm2hi_scalar} with CSE: (insn 16 15 17 2 (set (reg/v:RVVM2HI 137 [ output_var_0 ]) (if_then_else:RVVM2HI (unspec:RVVMF8BI [ (const_vector:RVVMF8BI repeat [ (const_int 1 [0x1]) ]) (reg:DI 146) (const_int 2 [0x2]) repeated x2 (const_int 0 [0]) (reg:SI 66 vl) (reg:SI 67 vtype) ] UNSPEC_VPREDICATE) (sign_extend:RVVM2HI (reg/v:RVVM1QI 136 [ op1 ])) (unspec:RVVM2HI [ (reg:SI 0 zero) ] UNSPEC_VUNDEF))) {pred_extendrvvm2hi_vf2} ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:32 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Ok. Let's hold on "RISC-V: Handle differences between XTheadvector and Vector" patch until you can reproduce the issue for me. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:29 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. The sext/zext issue is not related to xtheadvector-special patterns. I added !TARGET_XTHEADVECTOR to sext/zext patterns in "RISC-V: Handle differences between XTheadvector and Vector" That is caused by the vwmul pattern, but I cannot reproduce it right now. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:24 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. I don't see any patterns are possible CSE into sext/zext patterns: +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:21 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. "I didn't see theadvector-specific extension patterns. Could you show me?" They are all in the file thead-vector.md. For the sext/zext issue, perhaps I need some time to reproduce that optimization, but I can clearly remember it is related to vwmul. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:17 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. You mean which pattern optimized sext/vzext pattern? I didn't see theadvector-specific extension patterns. Could you show me? juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:14 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. To be specific, in CSE pass, the initial pattern will be optimized into the sext/zext pattern. ------------------------------------------------------------------ 发件人：joshua <cooper.joshua@linux.alibaba.com> 发送时间：2024年1月11日(星期四) 17:11 收件人："juzhe.zhong@rivai.ai"<juzhe.zhong@rivai.ai>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. sext/zext will be generated in O2 even without corresponding intrinsics. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:07 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ Add theadvector to the end of the enum. + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; Same. + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" You can use whole_reg_to_reg_move_p Btw, I review again : RISC-V: Handle differences between XTheadvector and Vector (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3713,7 +3744,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3734,7 +3765,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") Why do you add these !TARGERT_XTHEADVECRTOR ? juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-11 16:46 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (class th_extract): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (struct th_extract_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): * config/riscv/riscv-vector-builtins.h (enum required_ext): (struct function_group_info): * config/riscv/t-riscv: Add thead-vector-builtins-functions.def * config/riscv/thead-vector.md (@pred_mov_width<vlmem_op_attr><mode>): Add new patterns. (*pred_mov_width<vlmem_op_attr><mode>): Likewise. (@pred_store_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_load_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_store_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_load_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>): (@pred_th_extract<mode>): Likewise. (*pred_th_extract<mode>): Likewise. * config/riscv/thead-vector-builtins-functions.def: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 +++ .../riscv/riscv-vector-builtins-shapes.cc | 160 +++++++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + gcc/config/riscv/riscv-vector-builtins.cc | 70 +++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 253 ++++++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 +++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 +++++ 15 files changed, 1107 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 46f1a1da33e..3eba7943757 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2125,6 +2125,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2587,6 +2664,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> vsuxw_obj; +static CONSTEXPR const th_extract vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2878,6 +2986,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (vlb) +BASE (vlh) +BASE (vlw) +BASE (vlbu) +BASE (vlhu) +BASE (vlwu) +BASE (vsb) +BASE (vsh) +BASE (vsw) +BASE (vlsb) +BASE (vlsh) +BASE (vlsw) +BASE (vlsbu) +BASE (vlshu) +BASE (vlswu) +BASE (vssb) +BASE (vssh) +BASE (vssw) +BASE (vlxb) +BASE (vlxh) +BASE (vlxw) +BASE (vlxbu) +BASE (vlxhu) +BASE (vlxwu) +BASE (vsxb) +BASE (vsxh) +BASE (vsxw) +BASE (vsuxb) +BASE (vsuxh) +BASE (vsuxw) +BASE (vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 1122e3801a7..565a0311d2b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -299,6 +299,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const vlb; +extern const function_base *const vlh; +extern const function_base *const vlw; +extern const function_base *const vlbu; +extern const function_base *const vlhu; +extern const function_base *const vlwu; +extern const function_base *const vsb; +extern const function_base *const vsh; +extern const function_base *const vsw; +extern const function_base *const vlsb; +extern const function_base *const vlsh; +extern const function_base *const vlsw; +extern const function_base *const vlsbu; +extern const function_base *const vlshu; +extern const function_base *const vlswu; +extern const function_base *const vssb; +extern const function_base *const vssh; +extern const function_base *const vssw; +extern const function_base *const vlxb; +extern const function_base *const vlxh; +extern const function_base *const vlxw; +extern const function_base *const vlxbu; +extern const function_base *const vlxhu; +extern const function_base *const vlxwu; +extern const function_base *const vsxb; +extern const function_base *const vsxh; +extern const function_base *const vsxw; +extern const function_base *const vsuxb; +extern const function_base *const vsuxh; +extern const function_base *const vsuxw; +extern const function_base *const vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..8e90b17a94b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,146 @@ struct indexed_loadstore_def : public function_shape } }; +/* Add one function instance for GROUP, using operand suffix at index OI, + mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ +static void +build_th_loadstore (function_builder &b, const function_group_info &group, + unsigned int pred_idx, unsigned int vec_type_idx) +{ + auto_vec<tree, 5> argument_types; + function_instance function_instance (group.base_name, *group.base, + *group.shape, + group.ops_infos.types[vec_type_idx], + group.preds[pred_idx], &group.ops_infos); + tree return_type = group.ops_infos.ret.get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + b.allocate_argument_types (function_instance, argument_types); + b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + + tree type = builtin_types[group.ops_infos.types[vec_type_idx].index].vector; + if (strstr (group.base_name, "l") + && strstr (group.base_name, "u") + && !TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + if (strstr (group.base_name, "l") + && !strstr (group.base_name, "u") + && TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + machine_mode mode = TYPE_MODE (type); + int sew = GET_MODE_BITSIZE (GET_MODE_INNER (mode)); + if (strstr (group.base_name, "h") && sew == 8) + return; + + if (strstr (group.base_name, "w") && (sew == 8 || sew ==16)) + return; + + b.add_overloaded_function (function_instance, *group.shape); + b.add_unique_function (function_instance, (*group.shape), return_type, + argument_types); +} + +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +772,23 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1251,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1265,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..44b9fec1898 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -934,6 +934,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1481,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2672,38 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * index_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +2882,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..a8ee39a3cb2 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -114,6 +114,7 @@ static const unsigned int CP_WRITE_CSR = 1U << 5; enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ @@ -234,6 +235,8 @@ struct function_group_info { case VECTOR_EXT: return TARGET_VECTOR; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; case ZVBB_EXT: return TARGET_ZVBB; case ZVBB_OR_ZVKB_EXT: diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..fd3ba29bae9 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlhu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlwu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vsb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vlsb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlshu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlswu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vssb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vlxb, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxh, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxw, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxbu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxhu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxwu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vsxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..0f3700d9269 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,168 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..3c12c124597 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..30bef369375 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..3c8b5ccc16b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..b7c00404f18 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..17a53012acf --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..b187cfc852b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-11 12:05 ` joshua @ 2024-01-11 12:13 ` juzhe.zhong 2024-01-11 12:18 ` Re：Re：[PATCH " joshua 0 siblings, 1 reply; 133+ messages in thread From: juzhe.zhong @ 2024-01-11 12:13 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 72144 bytes --] I guess it is because of patterns from autovectorization pattern: ;; Use define_insn_and_split to define vsext.vf2/vzext.vf2 will help ;; to combine instructions as below: ;; vsext.vf2 + vsext.vf2 + vadd.vv ==> vwadd.vv (define_insn_and_split "<optab><v_double_trunc><mode>2" [(set (match_operand:VWEXTI 0 "register_operand" "=&vr") (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 1 "register_operand" "vr")))] "TARGET_VECTOR && can_create_pseudo_p ()" "#" "&& 1" [(const_int 0)] { insn_code icode = code_for_pred_vf2 (<CODE>, <MODE>mode); riscv_vector::emit_vlmax_insn (icode, riscv_vector::UNARY_OP, operands); DONE; } [(set_attr "type" "vext") (set_attr "mode" "<MODE>")]) juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 20:05 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. without CSE: (insn 16 15 17 2 (set (reg/v:RVVM2HI 137 [ output_var_0 ]) (if_then_else:RVVM2HI (unspec:RVVMF8BI [ (const_vector:RVVMF8BI repeat [ (const_int 1 [0x1]) ]) (reg:DI 146) (const_int 2 [0x2]) repeated x2 (const_int 0 [0]) (reg:SI 66 vl) (reg:SI 67 vtype) ] UNSPEC_VPREDICATE) (mult:RVVM2HI (sign_extend:RVVM2HI (reg/v:RVVM1QI 136 [ op1 ])) (sign_extend:RVVM2HI (vec_duplicate:RVVM1QI (reg:QI 147)))) (unspec:RVVM2HI [ (reg:SI 0 zero) ] UNSPEC_VUNDEF))) {pred_dual_widen_mulsrvvm2hi_scalar} with CSE: (insn 16 15 17 2 (set (reg/v:RVVM2HI 137 [ output_var_0 ]) (if_then_else:RVVM2HI (unspec:RVVMF8BI [ (const_vector:RVVMF8BI repeat [ (const_int 1 [0x1]) ]) (reg:DI 146) (const_int 2 [0x2]) repeated x2 (const_int 0 [0]) (reg:SI 66 vl) (reg:SI 67 vtype) ] UNSPEC_VPREDICATE) (sign_extend:RVVM2HI (reg/v:RVVM1QI 136 [ op1 ])) (unspec:RVVM2HI [ (reg:SI 0 zero) ] UNSPEC_VUNDEF))) {pred_extendrvvm2hi_vf2} ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:32 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Ok. Let's hold on "RISC-V: Handle differences between XTheadvector and Vector" patch until you can reproduce the issue for me. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:29 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. The sext/zext issue is not related to xtheadvector-special patterns. I added !TARGET_XTHEADVECTOR to sext/zext patterns in "RISC-V: Handle differences between XTheadvector and Vector" That is caused by the vwmul pattern, but I cannot reproduce it right now. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:24 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. I don't see any patterns are possible CSE into sext/zext patterns: +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:21 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. "I didn't see theadvector-specific extension patterns. Could you show me?" They are all in the file thead-vector.md. For the sext/zext issue, perhaps I need some time to reproduce that optimization, but I can clearly remember it is related to vwmul. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:17 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. You mean which pattern optimized sext/vzext pattern? I didn't see theadvector-specific extension patterns. Could you show me? juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:14 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. To be specific, in CSE pass, the initial pattern will be optimized into the sext/zext pattern. ------------------------------------------------------------------ 发件人：joshua <cooper.joshua@linux.alibaba.com> 发送时间：2024年1月11日(星期四) 17:11 收件人："juzhe.zhong@rivai.ai"<juzhe.zhong@rivai.ai>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. sext/zext will be generated in O2 even without corresponding intrinsics. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:07 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ Add theadvector to the end of the enum. + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; Same. + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" You can use whole_reg_to_reg_move_p Btw, I review again : RISC-V: Handle differences between XTheadvector and Vector (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3713,7 +3744,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3734,7 +3765,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") Why do you add these !TARGERT_XTHEADVECRTOR ? juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-11 16:46 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (class th_extract): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (struct th_extract_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): * config/riscv/riscv-vector-builtins.h (enum required_ext): (struct function_group_info): * config/riscv/t-riscv: Add thead-vector-builtins-functions.def * config/riscv/thead-vector.md (@pred_mov_width<vlmem_op_attr><mode>): Add new patterns. (*pred_mov_width<vlmem_op_attr><mode>): Likewise. (@pred_store_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_load_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_store_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_load_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>): (@pred_th_extract<mode>): Likewise. (*pred_th_extract<mode>): Likewise. * config/riscv/thead-vector-builtins-functions.def: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 +++ .../riscv/riscv-vector-builtins-shapes.cc | 160 +++++++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + gcc/config/riscv/riscv-vector-builtins.cc | 70 +++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 253 ++++++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 +++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 +++++ 15 files changed, 1107 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 46f1a1da33e..3eba7943757 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2125,6 +2125,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2587,6 +2664,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> vsuxw_obj; +static CONSTEXPR const th_extract vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2878,6 +2986,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (vlb) +BASE (vlh) +BASE (vlw) +BASE (vlbu) +BASE (vlhu) +BASE (vlwu) +BASE (vsb) +BASE (vsh) +BASE (vsw) +BASE (vlsb) +BASE (vlsh) +BASE (vlsw) +BASE (vlsbu) +BASE (vlshu) +BASE (vlswu) +BASE (vssb) +BASE (vssh) +BASE (vssw) +BASE (vlxb) +BASE (vlxh) +BASE (vlxw) +BASE (vlxbu) +BASE (vlxhu) +BASE (vlxwu) +BASE (vsxb) +BASE (vsxh) +BASE (vsxw) +BASE (vsuxb) +BASE (vsuxh) +BASE (vsuxw) +BASE (vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 1122e3801a7..565a0311d2b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -299,6 +299,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const vlb; +extern const function_base *const vlh; +extern const function_base *const vlw; +extern const function_base *const vlbu; +extern const function_base *const vlhu; +extern const function_base *const vlwu; +extern const function_base *const vsb; +extern const function_base *const vsh; +extern const function_base *const vsw; +extern const function_base *const vlsb; +extern const function_base *const vlsh; +extern const function_base *const vlsw; +extern const function_base *const vlsbu; +extern const function_base *const vlshu; +extern const function_base *const vlswu; +extern const function_base *const vssb; +extern const function_base *const vssh; +extern const function_base *const vssw; +extern const function_base *const vlxb; +extern const function_base *const vlxh; +extern const function_base *const vlxw; +extern const function_base *const vlxbu; +extern const function_base *const vlxhu; +extern const function_base *const vlxwu; +extern const function_base *const vsxb; +extern const function_base *const vsxh; +extern const function_base *const vsxw; +extern const function_base *const vsuxb; +extern const function_base *const vsuxh; +extern const function_base *const vsuxw; +extern const function_base *const vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..8e90b17a94b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,146 @@ struct indexed_loadstore_def : public function_shape } }; +/* Add one function instance for GROUP, using operand suffix at index OI, + mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ +static void +build_th_loadstore (function_builder &b, const function_group_info &group, + unsigned int pred_idx, unsigned int vec_type_idx) +{ + auto_vec<tree, 5> argument_types; + function_instance function_instance (group.base_name, *group.base, + *group.shape, + group.ops_infos.types[vec_type_idx], + group.preds[pred_idx], &group.ops_infos); + tree return_type = group.ops_infos.ret.get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + b.allocate_argument_types (function_instance, argument_types); + b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + + tree type = builtin_types[group.ops_infos.types[vec_type_idx].index].vector; + if (strstr (group.base_name, "l") + && strstr (group.base_name, "u") + && !TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + if (strstr (group.base_name, "l") + && !strstr (group.base_name, "u") + && TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + machine_mode mode = TYPE_MODE (type); + int sew = GET_MODE_BITSIZE (GET_MODE_INNER (mode)); + if (strstr (group.base_name, "h") && sew == 8) + return; + + if (strstr (group.base_name, "w") && (sew == 8 || sew ==16)) + return; + + b.add_overloaded_function (function_instance, *group.shape); + b.add_unique_function (function_instance, (*group.shape), return_type, + argument_types); +} + +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +772,23 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1251,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1265,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..44b9fec1898 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -934,6 +934,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1481,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2672,38 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * index_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +2882,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..a8ee39a3cb2 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -114,6 +114,7 @@ static const unsigned int CP_WRITE_CSR = 1U << 5; enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ @@ -234,6 +235,8 @@ struct function_group_info { case VECTOR_EXT: return TARGET_VECTOR; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; case ZVBB_EXT: return TARGET_ZVBB; case ZVBB_OR_ZVKB_EXT: diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..fd3ba29bae9 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlhu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlwu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vsb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vlsb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlshu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlswu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vssb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vlxb, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxh, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxw, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxbu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxhu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxwu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vsxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..0f3700d9269 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,168 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..3c12c124597 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..30bef369375 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..3c8b5ccc16b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..b7c00404f18 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..17a53012acf --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..b187cfc852b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-11 12:13 ` Re：[PATCH " juzhe.zhong @ 2024-01-11 12:18 ` joshua 2024-01-11 12:28 ` Re：[PATCH " juzhe.zhong 0 siblings, 1 reply; 133+ messages in thread From: joshua @ 2024-01-11 12:18 UTC (permalink / raw) To: juzhe.zhong, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu No, we have debugged and the code won't enter this autovec pattern. The root cause is CSE, in a very early pass. What I sent to you just now is the dump result which shows the difference. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 20:13 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. I guess it is because of patterns from autovectorization pattern: ;; Use define_insn_and_split to define vsext.vf2/vzext.vf2 will help ;; to combine instructions as below: ;; vsext.vf2 + vsext.vf2 + vadd.vv ==> vwadd.vv (define_insn_and_split "<optab><v_double_trunc><mode>2" [(set (match_operand:VWEXTI 0 "register_operand" "=&vr") (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 1 "register_operand" "vr")))] "TARGET_VECTOR && can_create_pseudo_p ()" "#" "&& 1" [(const_int 0)] { insn_code icode = code_for_pred_vf2 (<CODE>, <MODE>mode); riscv_vector::emit_vlmax_insn (icode, riscv_vector::UNARY_OP, operands); DONE; } [(set_attr "type" "vext") (set_attr "mode" "<MODE>")]) juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 20:05 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. without CSE: (insn 16 15 17 2 (set (reg/v:RVVM2HI 137 [ output_var_0 ]) (if_then_else:RVVM2HI (unspec:RVVMF8BI [ (const_vector:RVVMF8BI repeat [ (const_int 1 [0x1]) ]) (reg:DI 146) (const_int 2 [0x2]) repeated x2 (const_int 0 [0]) (reg:SI 66 vl) (reg:SI 67 vtype) ] UNSPEC_VPREDICATE) (mult:RVVM2HI (sign_extend:RVVM2HI (reg/v:RVVM1QI 136 [ op1 ])) (sign_extend:RVVM2HI (vec_duplicate:RVVM1QI (reg:QI 147)))) (unspec:RVVM2HI [ (reg:SI 0 zero) ] UNSPEC_VUNDEF))) {pred_dual_widen_mulsrvvm2hi_scalar} with CSE: (insn 16 15 17 2 (set (reg/v:RVVM2HI 137 [ output_var_0 ]) (if_then_else:RVVM2HI (unspec:RVVMF8BI [ (const_vector:RVVMF8BI repeat [ (const_int 1 [0x1]) ]) (reg:DI 146) (const_int 2 [0x2]) repeated x2 (const_int 0 [0]) (reg:SI 66 vl) (reg:SI 67 vtype) ] UNSPEC_VPREDICATE) (sign_extend:RVVM2HI (reg/v:RVVM1QI 136 [ op1 ])) (unspec:RVVM2HI [ (reg:SI 0 zero) ] UNSPEC_VUNDEF))) {pred_extendrvvm2hi_vf2} ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:32 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Ok. Let's hold on "RISC-V: Handle differences between XTheadvector and Vector" patch until you can reproduce the issue for me. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:29 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. The sext/zext issue is not related to xtheadvector-special patterns. I added !TARGET_XTHEADVECTOR to sext/zext patterns in "RISC-V: Handle differences between XTheadvector and Vector" That is caused by the vwmul pattern, but I cannot reproduce it right now. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:24 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. I don't see any patterns are possible CSE into sext/zext patterns: +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:21 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. "I didn't see theadvector-specific extension patterns. Could you show me?" They are all in the file thead-vector.md. For the sext/zext issue, perhaps I need some time to reproduce that optimization, but I can clearly remember it is related to vwmul. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:17 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. You mean which pattern optimized sext/vzext pattern? I didn't see theadvector-specific extension patterns. Could you show me? juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:14 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. To be specific, in CSE pass, the initial pattern will be optimized into the sext/zext pattern. ------------------------------------------------------------------ 发件人：joshua <cooper.joshua@linux.alibaba.com> 发送时间：2024年1月11日(星期四) 17:11 收件人："juzhe.zhong@rivai.ai"<juzhe.zhong@rivai.ai>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. sext/zext will be generated in O2 even without corresponding intrinsics. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:07 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ Add theadvector to the end of the enum. + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; Same. + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" You can use whole_reg_to_reg_move_p Btw, I review again : RISC-V: Handle differences between XTheadvector and Vector (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3713,7 +3744,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3734,7 +3765,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") Why do you add these !TARGERT_XTHEADVECRTOR ? juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-11 16:46 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (class th_extract): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (struct th_extract_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): * config/riscv/riscv-vector-builtins.h (enum required_ext): (struct function_group_info): * config/riscv/t-riscv: Add thead-vector-builtins-functions.def * config/riscv/thead-vector.md (@pred_mov_width<vlmem_op_attr><mode>): Add new patterns. (*pred_mov_width<vlmem_op_attr><mode>): Likewise. (@pred_store_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_load_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_store_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_load_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>): (@pred_th_extract<mode>): Likewise. (*pred_th_extract<mode>): Likewise. * config/riscv/thead-vector-builtins-functions.def: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 +++ .../riscv/riscv-vector-builtins-shapes.cc | 160 +++++++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + gcc/config/riscv/riscv-vector-builtins.cc | 70 +++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 253 ++++++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 +++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 +++++ 15 files changed, 1107 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 46f1a1da33e..3eba7943757 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2125,6 +2125,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2587,6 +2664,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> vsuxw_obj; +static CONSTEXPR const th_extract vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2878,6 +2986,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (vlb) +BASE (vlh) +BASE (vlw) +BASE (vlbu) +BASE (vlhu) +BASE (vlwu) +BASE (vsb) +BASE (vsh) +BASE (vsw) +BASE (vlsb) +BASE (vlsh) +BASE (vlsw) +BASE (vlsbu) +BASE (vlshu) +BASE (vlswu) +BASE (vssb) +BASE (vssh) +BASE (vssw) +BASE (vlxb) +BASE (vlxh) +BASE (vlxw) +BASE (vlxbu) +BASE (vlxhu) +BASE (vlxwu) +BASE (vsxb) +BASE (vsxh) +BASE (vsxw) +BASE (vsuxb) +BASE (vsuxh) +BASE (vsuxw) +BASE (vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 1122e3801a7..565a0311d2b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -299,6 +299,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const vlb; +extern const function_base *const vlh; +extern const function_base *const vlw; +extern const function_base *const vlbu; +extern const function_base *const vlhu; +extern const function_base *const vlwu; +extern const function_base *const vsb; +extern const function_base *const vsh; +extern const function_base *const vsw; +extern const function_base *const vlsb; +extern const function_base *const vlsh; +extern const function_base *const vlsw; +extern const function_base *const vlsbu; +extern const function_base *const vlshu; +extern const function_base *const vlswu; +extern const function_base *const vssb; +extern const function_base *const vssh; +extern const function_base *const vssw; +extern const function_base *const vlxb; +extern const function_base *const vlxh; +extern const function_base *const vlxw; +extern const function_base *const vlxbu; +extern const function_base *const vlxhu; +extern const function_base *const vlxwu; +extern const function_base *const vsxb; +extern const function_base *const vsxh; +extern const function_base *const vsxw; +extern const function_base *const vsuxb; +extern const function_base *const vsuxh; +extern const function_base *const vsuxw; +extern const function_base *const vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..8e90b17a94b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,146 @@ struct indexed_loadstore_def : public function_shape } }; +/* Add one function instance for GROUP, using operand suffix at index OI, + mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ +static void +build_th_loadstore (function_builder &b, const function_group_info &group, + unsigned int pred_idx, unsigned int vec_type_idx) +{ + auto_vec<tree, 5> argument_types; + function_instance function_instance (group.base_name, *group.base, + *group.shape, + group.ops_infos.types[vec_type_idx], + group.preds[pred_idx], &group.ops_infos); + tree return_type = group.ops_infos.ret.get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + b.allocate_argument_types (function_instance, argument_types); + b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + + tree type = builtin_types[group.ops_infos.types[vec_type_idx].index].vector; + if (strstr (group.base_name, "l") + && strstr (group.base_name, "u") + && !TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + if (strstr (group.base_name, "l") + && !strstr (group.base_name, "u") + && TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + machine_mode mode = TYPE_MODE (type); + int sew = GET_MODE_BITSIZE (GET_MODE_INNER (mode)); + if (strstr (group.base_name, "h") && sew == 8) + return; + + if (strstr (group.base_name, "w") && (sew == 8 || sew ==16)) + return; + + b.add_overloaded_function (function_instance, *group.shape); + b.add_unique_function (function_instance, (*group.shape), return_type, + argument_types); +} + +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +772,23 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1251,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1265,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..44b9fec1898 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -934,6 +934,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1481,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2672,38 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * index_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +2882,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..a8ee39a3cb2 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -114,6 +114,7 @@ static const unsigned int CP_WRITE_CSR = 1U << 5; enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ @@ -234,6 +235,8 @@ struct function_group_info { case VECTOR_EXT: return TARGET_VECTOR; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; case ZVBB_EXT: return TARGET_ZVBB; case ZVBB_OR_ZVKB_EXT: diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..fd3ba29bae9 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlhu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlwu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vsb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vlsb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlshu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlswu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vssb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vlxb, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxh, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxw, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxbu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxhu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxwu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vsxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..0f3700d9269 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,168 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..3c12c124597 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..30bef369375 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..3c8b5ccc16b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..b7c00404f18 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..17a53012acf --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..b187cfc852b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-11 12:18 ` Re：Re：[PATCH " joshua @ 2024-01-11 12:28 ` juzhe.zhong 2024-01-11 12:31 ` Re：Re：[PATCH " joshua 0 siblings, 1 reply; 133+ messages in thread From: juzhe.zhong @ 2024-01-11 12:28 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 73492 bytes --] (vec_duplicate:RVVM1QI (reg:QI 147)))) Find the RTL define pseudo 147 to me. I guess it is 1. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 20:18 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. No, we have debugged and the code won't enter this autovec pattern. The root cause is CSE, in a very early pass. What I sent to you just now is the dump result which shows the difference. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 20:13 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. I guess it is because of patterns from autovectorization pattern: ;; Use define_insn_and_split to define vsext.vf2/vzext.vf2 will help ;; to combine instructions as below: ;; vsext.vf2 + vsext.vf2 + vadd.vv ==> vwadd.vv (define_insn_and_split "<optab><v_double_trunc><mode>2" [(set (match_operand:VWEXTI 0 "register_operand" "=&vr") (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 1 "register_operand" "vr")))] "TARGET_VECTOR && can_create_pseudo_p ()" "#" "&& 1" [(const_int 0)] { insn_code icode = code_for_pred_vf2 (<CODE>, <MODE>mode); riscv_vector::emit_vlmax_insn (icode, riscv_vector::UNARY_OP, operands); DONE; } [(set_attr "type" "vext") (set_attr "mode" "<MODE>")]) juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 20:05 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. without CSE: (insn 16 15 17 2 (set (reg/v:RVVM2HI 137 [ output_var_0 ]) (if_then_else:RVVM2HI (unspec:RVVMF8BI [ (const_vector:RVVMF8BI repeat [ (const_int 1 [0x1]) ]) (reg:DI 146) (const_int 2 [0x2]) repeated x2 (const_int 0 [0]) (reg:SI 66 vl) (reg:SI 67 vtype) ] UNSPEC_VPREDICATE) (mult:RVVM2HI (sign_extend:RVVM2HI (reg/v:RVVM1QI 136 [ op1 ])) (sign_extend:RVVM2HI (vec_duplicate:RVVM1QI (reg:QI 147)))) (unspec:RVVM2HI [ (reg:SI 0 zero) ] UNSPEC_VUNDEF))) {pred_dual_widen_mulsrvvm2hi_scalar} with CSE: (insn 16 15 17 2 (set (reg/v:RVVM2HI 137 [ output_var_0 ]) (if_then_else:RVVM2HI (unspec:RVVMF8BI [ (const_vector:RVVMF8BI repeat [ (const_int 1 [0x1]) ]) (reg:DI 146) (const_int 2 [0x2]) repeated x2 (const_int 0 [0]) (reg:SI 66 vl) (reg:SI 67 vtype) ] UNSPEC_VPREDICATE) (sign_extend:RVVM2HI (reg/v:RVVM1QI 136 [ op1 ])) (unspec:RVVM2HI [ (reg:SI 0 zero) ] UNSPEC_VUNDEF))) {pred_extendrvvm2hi_vf2} ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:32 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Ok. Let's hold on "RISC-V: Handle differences between XTheadvector and Vector" patch until you can reproduce the issue for me. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:29 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. The sext/zext issue is not related to xtheadvector-special patterns. I added !TARGET_XTHEADVECTOR to sext/zext patterns in "RISC-V: Handle differences between XTheadvector and Vector" That is caused by the vwmul pattern, but I cannot reproduce it right now. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:24 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. I don't see any patterns are possible CSE into sext/zext patterns: +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:21 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. "I didn't see theadvector-specific extension patterns. Could you show me?" They are all in the file thead-vector.md. For the sext/zext issue, perhaps I need some time to reproduce that optimization, but I can clearly remember it is related to vwmul. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:17 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. You mean which pattern optimized sext/vzext pattern? I didn't see theadvector-specific extension patterns. Could you show me? juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:14 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. To be specific, in CSE pass, the initial pattern will be optimized into the sext/zext pattern. ------------------------------------------------------------------ 发件人：joshua <cooper.joshua@linux.alibaba.com> 发送时间：2024年1月11日(星期四) 17:11 收件人："juzhe.zhong@rivai.ai"<juzhe.zhong@rivai.ai>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. sext/zext will be generated in O2 even without corresponding intrinsics. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:07 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ Add theadvector to the end of the enum. + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; Same. + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" You can use whole_reg_to_reg_move_p Btw, I review again : RISC-V: Handle differences between XTheadvector and Vector (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3713,7 +3744,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3734,7 +3765,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") Why do you add these !TARGERT_XTHEADVECRTOR ? juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-11 16:46 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (class th_extract): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (struct th_extract_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): * config/riscv/riscv-vector-builtins.h (enum required_ext): (struct function_group_info): * config/riscv/t-riscv: Add thead-vector-builtins-functions.def * config/riscv/thead-vector.md (@pred_mov_width<vlmem_op_attr><mode>): Add new patterns. (*pred_mov_width<vlmem_op_attr><mode>): Likewise. (@pred_store_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_load_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_store_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_load_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>): (@pred_th_extract<mode>): Likewise. (*pred_th_extract<mode>): Likewise. * config/riscv/thead-vector-builtins-functions.def: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 +++ .../riscv/riscv-vector-builtins-shapes.cc | 160 +++++++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + gcc/config/riscv/riscv-vector-builtins.cc | 70 +++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 253 ++++++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 +++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 +++++ 15 files changed, 1107 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 46f1a1da33e..3eba7943757 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2125,6 +2125,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2587,6 +2664,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> vsuxw_obj; +static CONSTEXPR const th_extract vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2878,6 +2986,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (vlb) +BASE (vlh) +BASE (vlw) +BASE (vlbu) +BASE (vlhu) +BASE (vlwu) +BASE (vsb) +BASE (vsh) +BASE (vsw) +BASE (vlsb) +BASE (vlsh) +BASE (vlsw) +BASE (vlsbu) +BASE (vlshu) +BASE (vlswu) +BASE (vssb) +BASE (vssh) +BASE (vssw) +BASE (vlxb) +BASE (vlxh) +BASE (vlxw) +BASE (vlxbu) +BASE (vlxhu) +BASE (vlxwu) +BASE (vsxb) +BASE (vsxh) +BASE (vsxw) +BASE (vsuxb) +BASE (vsuxh) +BASE (vsuxw) +BASE (vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 1122e3801a7..565a0311d2b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -299,6 +299,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const vlb; +extern const function_base *const vlh; +extern const function_base *const vlw; +extern const function_base *const vlbu; +extern const function_base *const vlhu; +extern const function_base *const vlwu; +extern const function_base *const vsb; +extern const function_base *const vsh; +extern const function_base *const vsw; +extern const function_base *const vlsb; +extern const function_base *const vlsh; +extern const function_base *const vlsw; +extern const function_base *const vlsbu; +extern const function_base *const vlshu; +extern const function_base *const vlswu; +extern const function_base *const vssb; +extern const function_base *const vssh; +extern const function_base *const vssw; +extern const function_base *const vlxb; +extern const function_base *const vlxh; +extern const function_base *const vlxw; +extern const function_base *const vlxbu; +extern const function_base *const vlxhu; +extern const function_base *const vlxwu; +extern const function_base *const vsxb; +extern const function_base *const vsxh; +extern const function_base *const vsxw; +extern const function_base *const vsuxb; +extern const function_base *const vsuxh; +extern const function_base *const vsuxw; +extern const function_base *const vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..8e90b17a94b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,146 @@ struct indexed_loadstore_def : public function_shape } }; +/* Add one function instance for GROUP, using operand suffix at index OI, + mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ +static void +build_th_loadstore (function_builder &b, const function_group_info &group, + unsigned int pred_idx, unsigned int vec_type_idx) +{ + auto_vec<tree, 5> argument_types; + function_instance function_instance (group.base_name, *group.base, + *group.shape, + group.ops_infos.types[vec_type_idx], + group.preds[pred_idx], &group.ops_infos); + tree return_type = group.ops_infos.ret.get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + b.allocate_argument_types (function_instance, argument_types); + b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + + tree type = builtin_types[group.ops_infos.types[vec_type_idx].index].vector; + if (strstr (group.base_name, "l") + && strstr (group.base_name, "u") + && !TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + if (strstr (group.base_name, "l") + && !strstr (group.base_name, "u") + && TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + machine_mode mode = TYPE_MODE (type); + int sew = GET_MODE_BITSIZE (GET_MODE_INNER (mode)); + if (strstr (group.base_name, "h") && sew == 8) + return; + + if (strstr (group.base_name, "w") && (sew == 8 || sew ==16)) + return; + + b.add_overloaded_function (function_instance, *group.shape); + b.add_unique_function (function_instance, (*group.shape), return_type, + argument_types); +} + +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +772,23 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1251,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1265,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..44b9fec1898 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -934,6 +934,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1481,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2672,38 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * index_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +2882,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..a8ee39a3cb2 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -114,6 +114,7 @@ static const unsigned int CP_WRITE_CSR = 1U << 5; enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ @@ -234,6 +235,8 @@ struct function_group_info { case VECTOR_EXT: return TARGET_VECTOR; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; case ZVBB_EXT: return TARGET_ZVBB; case ZVBB_OR_ZVKB_EXT: diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..fd3ba29bae9 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlhu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlwu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vsb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vlsb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlshu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlswu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vssb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vlxb, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxh, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxw, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxbu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxhu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxwu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vsxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..0f3700d9269 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,168 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..3c12c124597 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..30bef369375 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..3c8b5ccc16b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..b7c00404f18 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..17a53012acf --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..b187cfc852b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-11 12:28 ` Re：[PATCH " juzhe.zhong @ 2024-01-11 12:31 ` joshua 2024-01-11 12:33 ` Re：[PATCH " juzhe.zhong 0 siblings, 1 reply; 133+ messages in thread From: joshua @ 2024-01-11 12:31 UTC (permalink / raw) To: juzhe.zhong, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu Yes. (insn 14 13 15 2 (set (reg:QI 147) (const_int 1 [0x1])) {*movqi_internal} (nil)) ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 20:28 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. (vec_duplicate:RVVM1QI (reg:QI 147)))) Find the RTL define pseudo 147 to me. I guess it is 1. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 20:18 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. No, we have debugged and the code won't enter this autovec pattern. The root cause is CSE, in a very early pass. What I sent to you just now is the dump result which shows the difference. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 20:13 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. I guess it is because of patterns from autovectorization pattern: ;; Use define_insn_and_split to define vsext.vf2/vzext.vf2 will help ;; to combine instructions as below: ;; vsext.vf2 + vsext.vf2 + vadd.vv ==> vwadd.vv (define_insn_and_split "<optab><v_double_trunc><mode>2" [(set (match_operand:VWEXTI 0 "register_operand" "=&vr") (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 1 "register_operand" "vr")))] "TARGET_VECTOR && can_create_pseudo_p ()" "#" "&& 1" [(const_int 0)] { insn_code icode = code_for_pred_vf2 (<CODE>, <MODE>mode); riscv_vector::emit_vlmax_insn (icode, riscv_vector::UNARY_OP, operands); DONE; } [(set_attr "type" "vext") (set_attr "mode" "<MODE>")]) juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 20:05 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. without CSE: (insn 16 15 17 2 (set (reg/v:RVVM2HI 137 [ output_var_0 ]) (if_then_else:RVVM2HI (unspec:RVVMF8BI [ (const_vector:RVVMF8BI repeat [ (const_int 1 [0x1]) ]) (reg:DI 146) (const_int 2 [0x2]) repeated x2 (const_int 0 [0]) (reg:SI 66 vl) (reg:SI 67 vtype) ] UNSPEC_VPREDICATE) (mult:RVVM2HI (sign_extend:RVVM2HI (reg/v:RVVM1QI 136 [ op1 ])) (sign_extend:RVVM2HI (vec_duplicate:RVVM1QI (reg:QI 147)))) (unspec:RVVM2HI [ (reg:SI 0 zero) ] UNSPEC_VUNDEF))) {pred_dual_widen_mulsrvvm2hi_scalar} with CSE: (insn 16 15 17 2 (set (reg/v:RVVM2HI 137 [ output_var_0 ]) (if_then_else:RVVM2HI (unspec:RVVMF8BI [ (const_vector:RVVMF8BI repeat [ (const_int 1 [0x1]) ]) (reg:DI 146) (const_int 2 [0x2]) repeated x2 (const_int 0 [0]) (reg:SI 66 vl) (reg:SI 67 vtype) ] UNSPEC_VPREDICATE) (sign_extend:RVVM2HI (reg/v:RVVM1QI 136 [ op1 ])) (unspec:RVVM2HI [ (reg:SI 0 zero) ] UNSPEC_VUNDEF))) {pred_extendrvvm2hi_vf2} ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:32 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Ok. Let's hold on "RISC-V: Handle differences between XTheadvector and Vector" patch until you can reproduce the issue for me. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:29 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. The sext/zext issue is not related to xtheadvector-special patterns. I added !TARGET_XTHEADVECTOR to sext/zext patterns in "RISC-V: Handle differences between XTheadvector and Vector" That is caused by the vwmul pattern, but I cannot reproduce it right now. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:24 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. I don't see any patterns are possible CSE into sext/zext patterns: +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:21 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. "I didn't see theadvector-specific extension patterns. Could you show me?" They are all in the file thead-vector.md. For the sext/zext issue, perhaps I need some time to reproduce that optimization, but I can clearly remember it is related to vwmul. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:17 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. You mean which pattern optimized sext/vzext pattern? I didn't see theadvector-specific extension patterns. Could you show me? juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:14 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. To be specific, in CSE pass, the initial pattern will be optimized into the sext/zext pattern. ------------------------------------------------------------------ 发件人：joshua <cooper.joshua@linux.alibaba.com> 发送时间：2024年1月11日(星期四) 17:11 收件人："juzhe.zhong@rivai.ai"<juzhe.zhong@rivai.ai>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. sext/zext will be generated in O2 even without corresponding intrinsics. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:07 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ Add theadvector to the end of the enum. + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; Same. + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" You can use whole_reg_to_reg_move_p Btw, I review again : RISC-V: Handle differences between XTheadvector and Vector (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3713,7 +3744,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3734,7 +3765,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") Why do you add these !TARGERT_XTHEADVECRTOR ? juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-11 16:46 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (class th_extract): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (struct th_extract_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): * config/riscv/riscv-vector-builtins.h (enum required_ext): (struct function_group_info): * config/riscv/t-riscv: Add thead-vector-builtins-functions.def * config/riscv/thead-vector.md (@pred_mov_width<vlmem_op_attr><mode>): Add new patterns. (*pred_mov_width<vlmem_op_attr><mode>): Likewise. (@pred_store_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_load_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_store_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_load_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>): (@pred_th_extract<mode>): Likewise. (*pred_th_extract<mode>): Likewise. * config/riscv/thead-vector-builtins-functions.def: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 +++ .../riscv/riscv-vector-builtins-shapes.cc | 160 +++++++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + gcc/config/riscv/riscv-vector-builtins.cc | 70 +++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 253 ++++++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 +++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 +++++ 15 files changed, 1107 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 46f1a1da33e..3eba7943757 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2125,6 +2125,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2587,6 +2664,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> vsuxw_obj; +static CONSTEXPR const th_extract vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2878,6 +2986,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (vlb) +BASE (vlh) +BASE (vlw) +BASE (vlbu) +BASE (vlhu) +BASE (vlwu) +BASE (vsb) +BASE (vsh) +BASE (vsw) +BASE (vlsb) +BASE (vlsh) +BASE (vlsw) +BASE (vlsbu) +BASE (vlshu) +BASE (vlswu) +BASE (vssb) +BASE (vssh) +BASE (vssw) +BASE (vlxb) +BASE (vlxh) +BASE (vlxw) +BASE (vlxbu) +BASE (vlxhu) +BASE (vlxwu) +BASE (vsxb) +BASE (vsxh) +BASE (vsxw) +BASE (vsuxb) +BASE (vsuxh) +BASE (vsuxw) +BASE (vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 1122e3801a7..565a0311d2b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -299,6 +299,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const vlb; +extern const function_base *const vlh; +extern const function_base *const vlw; +extern const function_base *const vlbu; +extern const function_base *const vlhu; +extern const function_base *const vlwu; +extern const function_base *const vsb; +extern const function_base *const vsh; +extern const function_base *const vsw; +extern const function_base *const vlsb; +extern const function_base *const vlsh; +extern const function_base *const vlsw; +extern const function_base *const vlsbu; +extern const function_base *const vlshu; +extern const function_base *const vlswu; +extern const function_base *const vssb; +extern const function_base *const vssh; +extern const function_base *const vssw; +extern const function_base *const vlxb; +extern const function_base *const vlxh; +extern const function_base *const vlxw; +extern const function_base *const vlxbu; +extern const function_base *const vlxhu; +extern const function_base *const vlxwu; +extern const function_base *const vsxb; +extern const function_base *const vsxh; +extern const function_base *const vsxw; +extern const function_base *const vsuxb; +extern const function_base *const vsuxh; +extern const function_base *const vsuxw; +extern const function_base *const vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..8e90b17a94b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,146 @@ struct indexed_loadstore_def : public function_shape } }; +/* Add one function instance for GROUP, using operand suffix at index OI, + mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ +static void +build_th_loadstore (function_builder &b, const function_group_info &group, + unsigned int pred_idx, unsigned int vec_type_idx) +{ + auto_vec<tree, 5> argument_types; + function_instance function_instance (group.base_name, *group.base, + *group.shape, + group.ops_infos.types[vec_type_idx], + group.preds[pred_idx], &group.ops_infos); + tree return_type = group.ops_infos.ret.get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + b.allocate_argument_types (function_instance, argument_types); + b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + + tree type = builtin_types[group.ops_infos.types[vec_type_idx].index].vector; + if (strstr (group.base_name, "l") + && strstr (group.base_name, "u") + && !TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + if (strstr (group.base_name, "l") + && !strstr (group.base_name, "u") + && TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + machine_mode mode = TYPE_MODE (type); + int sew = GET_MODE_BITSIZE (GET_MODE_INNER (mode)); + if (strstr (group.base_name, "h") && sew == 8) + return; + + if (strstr (group.base_name, "w") && (sew == 8 || sew ==16)) + return; + + b.add_overloaded_function (function_instance, *group.shape); + b.add_unique_function (function_instance, (*group.shape), return_type, + argument_types); +} + +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +772,23 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1251,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1265,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..44b9fec1898 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -934,6 +934,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1481,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2672,38 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * index_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +2882,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..a8ee39a3cb2 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -114,6 +114,7 @@ static const unsigned int CP_WRITE_CSR = 1U << 5; enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ @@ -234,6 +235,8 @@ struct function_group_info { case VECTOR_EXT: return TARGET_VECTOR; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; case ZVBB_EXT: return TARGET_ZVBB; case ZVBB_OR_ZVKB_EXT: diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..fd3ba29bae9 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlhu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlwu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vsb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vlsb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlshu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlswu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vssb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vlxb, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxh, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxw, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxbu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxhu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxwu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vsxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..0f3700d9269 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,168 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..3c12c124597 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..30bef369375 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..3c8b5ccc16b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..b7c00404f18 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..17a53012acf --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..b187cfc852b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-11 12:31 ` Re：Re：[PATCH " joshua @ 2024-01-11 12:33 ` juzhe.zhong 2024-01-11 12:36 ` Re：Re：[PATCH " joshua 0 siblings, 1 reply; 133+ messages in thread From: juzhe.zhong @ 2024-01-11 12:33 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 74411 bytes --] Does theadvector has extension instructions ? Show me the pattern. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 20:31 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Yes. (insn 14 13 15 2 (set (reg:QI 147) (const_int 1 [0x1])) {*movqi_internal} (nil)) ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 20:28 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. (vec_duplicate:RVVM1QI (reg:QI 147)))) Find the RTL define pseudo 147 to me. I guess it is 1. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 20:18 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. No, we have debugged and the code won't enter this autovec pattern. The root cause is CSE, in a very early pass. What I sent to you just now is the dump result which shows the difference. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 20:13 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. I guess it is because of patterns from autovectorization pattern: ;; Use define_insn_and_split to define vsext.vf2/vzext.vf2 will help ;; to combine instructions as below: ;; vsext.vf2 + vsext.vf2 + vadd.vv ==> vwadd.vv (define_insn_and_split "<optab><v_double_trunc><mode>2" [(set (match_operand:VWEXTI 0 "register_operand" "=&vr") (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 1 "register_operand" "vr")))] "TARGET_VECTOR && can_create_pseudo_p ()" "#" "&& 1" [(const_int 0)] { insn_code icode = code_for_pred_vf2 (<CODE>, <MODE>mode); riscv_vector::emit_vlmax_insn (icode, riscv_vector::UNARY_OP, operands); DONE; } [(set_attr "type" "vext") (set_attr "mode" "<MODE>")]) juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 20:05 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. without CSE: (insn 16 15 17 2 (set (reg/v:RVVM2HI 137 [ output_var_0 ]) (if_then_else:RVVM2HI (unspec:RVVMF8BI [ (const_vector:RVVMF8BI repeat [ (const_int 1 [0x1]) ]) (reg:DI 146) (const_int 2 [0x2]) repeated x2 (const_int 0 [0]) (reg:SI 66 vl) (reg:SI 67 vtype) ] UNSPEC_VPREDICATE) (mult:RVVM2HI (sign_extend:RVVM2HI (reg/v:RVVM1QI 136 [ op1 ])) (sign_extend:RVVM2HI (vec_duplicate:RVVM1QI (reg:QI 147)))) (unspec:RVVM2HI [ (reg:SI 0 zero) ] UNSPEC_VUNDEF))) {pred_dual_widen_mulsrvvm2hi_scalar} with CSE: (insn 16 15 17 2 (set (reg/v:RVVM2HI 137 [ output_var_0 ]) (if_then_else:RVVM2HI (unspec:RVVMF8BI [ (const_vector:RVVMF8BI repeat [ (const_int 1 [0x1]) ]) (reg:DI 146) (const_int 2 [0x2]) repeated x2 (const_int 0 [0]) (reg:SI 66 vl) (reg:SI 67 vtype) ] UNSPEC_VPREDICATE) (sign_extend:RVVM2HI (reg/v:RVVM1QI 136 [ op1 ])) (unspec:RVVM2HI [ (reg:SI 0 zero) ] UNSPEC_VUNDEF))) {pred_extendrvvm2hi_vf2} ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:32 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Ok. Let's hold on "RISC-V: Handle differences between XTheadvector and Vector" patch until you can reproduce the issue for me. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:29 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. The sext/zext issue is not related to xtheadvector-special patterns. I added !TARGET_XTHEADVECTOR to sext/zext patterns in "RISC-V: Handle differences between XTheadvector and Vector" That is caused by the vwmul pattern, but I cannot reproduce it right now. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:24 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. I don't see any patterns are possible CSE into sext/zext patterns: +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:21 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. "I didn't see theadvector-specific extension patterns. Could you show me?" They are all in the file thead-vector.md. For the sext/zext issue, perhaps I need some time to reproduce that optimization, but I can clearly remember it is related to vwmul. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:17 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. You mean which pattern optimized sext/vzext pattern? I didn't see theadvector-specific extension patterns. Could you show me? juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:14 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. To be specific, in CSE pass, the initial pattern will be optimized into the sext/zext pattern. ------------------------------------------------------------------ 发件人：joshua <cooper.joshua@linux.alibaba.com> 发送时间：2024年1月11日(星期四) 17:11 收件人："juzhe.zhong@rivai.ai"<juzhe.zhong@rivai.ai>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. sext/zext will be generated in O2 even without corresponding intrinsics. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:07 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ Add theadvector to the end of the enum. + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; Same. + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" You can use whole_reg_to_reg_move_p Btw, I review again : RISC-V: Handle differences between XTheadvector and Vector (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3713,7 +3744,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3734,7 +3765,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") Why do you add these !TARGERT_XTHEADVECRTOR ? juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-11 16:46 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (class th_extract): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (struct th_extract_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): * config/riscv/riscv-vector-builtins.h (enum required_ext): (struct function_group_info): * config/riscv/t-riscv: Add thead-vector-builtins-functions.def * config/riscv/thead-vector.md (@pred_mov_width<vlmem_op_attr><mode>): Add new patterns. (*pred_mov_width<vlmem_op_attr><mode>): Likewise. (@pred_store_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_load_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_store_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_load_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>): (@pred_th_extract<mode>): Likewise. (*pred_th_extract<mode>): Likewise. * config/riscv/thead-vector-builtins-functions.def: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 +++ .../riscv/riscv-vector-builtins-shapes.cc | 160 +++++++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + gcc/config/riscv/riscv-vector-builtins.cc | 70 +++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 253 ++++++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 +++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 +++++ 15 files changed, 1107 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 46f1a1da33e..3eba7943757 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2125,6 +2125,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2587,6 +2664,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> vsuxw_obj; +static CONSTEXPR const th_extract vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2878,6 +2986,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (vlb) +BASE (vlh) +BASE (vlw) +BASE (vlbu) +BASE (vlhu) +BASE (vlwu) +BASE (vsb) +BASE (vsh) +BASE (vsw) +BASE (vlsb) +BASE (vlsh) +BASE (vlsw) +BASE (vlsbu) +BASE (vlshu) +BASE (vlswu) +BASE (vssb) +BASE (vssh) +BASE (vssw) +BASE (vlxb) +BASE (vlxh) +BASE (vlxw) +BASE (vlxbu) +BASE (vlxhu) +BASE (vlxwu) +BASE (vsxb) +BASE (vsxh) +BASE (vsxw) +BASE (vsuxb) +BASE (vsuxh) +BASE (vsuxw) +BASE (vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 1122e3801a7..565a0311d2b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -299,6 +299,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const vlb; +extern const function_base *const vlh; +extern const function_base *const vlw; +extern const function_base *const vlbu; +extern const function_base *const vlhu; +extern const function_base *const vlwu; +extern const function_base *const vsb; +extern const function_base *const vsh; +extern const function_base *const vsw; +extern const function_base *const vlsb; +extern const function_base *const vlsh; +extern const function_base *const vlsw; +extern const function_base *const vlsbu; +extern const function_base *const vlshu; +extern const function_base *const vlswu; +extern const function_base *const vssb; +extern const function_base *const vssh; +extern const function_base *const vssw; +extern const function_base *const vlxb; +extern const function_base *const vlxh; +extern const function_base *const vlxw; +extern const function_base *const vlxbu; +extern const function_base *const vlxhu; +extern const function_base *const vlxwu; +extern const function_base *const vsxb; +extern const function_base *const vsxh; +extern const function_base *const vsxw; +extern const function_base *const vsuxb; +extern const function_base *const vsuxh; +extern const function_base *const vsuxw; +extern const function_base *const vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..8e90b17a94b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,146 @@ struct indexed_loadstore_def : public function_shape } }; +/* Add one function instance for GROUP, using operand suffix at index OI, + mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ +static void +build_th_loadstore (function_builder &b, const function_group_info &group, + unsigned int pred_idx, unsigned int vec_type_idx) +{ + auto_vec<tree, 5> argument_types; + function_instance function_instance (group.base_name, *group.base, + *group.shape, + group.ops_infos.types[vec_type_idx], + group.preds[pred_idx], &group.ops_infos); + tree return_type = group.ops_infos.ret.get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + b.allocate_argument_types (function_instance, argument_types); + b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + + tree type = builtin_types[group.ops_infos.types[vec_type_idx].index].vector; + if (strstr (group.base_name, "l") + && strstr (group.base_name, "u") + && !TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + if (strstr (group.base_name, "l") + && !strstr (group.base_name, "u") + && TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + machine_mode mode = TYPE_MODE (type); + int sew = GET_MODE_BITSIZE (GET_MODE_INNER (mode)); + if (strstr (group.base_name, "h") && sew == 8) + return; + + if (strstr (group.base_name, "w") && (sew == 8 || sew ==16)) + return; + + b.add_overloaded_function (function_instance, *group.shape); + b.add_unique_function (function_instance, (*group.shape), return_type, + argument_types); +} + +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +772,23 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1251,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1265,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..44b9fec1898 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -934,6 +934,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1481,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2672,38 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * index_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +2882,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..a8ee39a3cb2 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -114,6 +114,7 @@ static const unsigned int CP_WRITE_CSR = 1U << 5; enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ @@ -234,6 +235,8 @@ struct function_group_info { case VECTOR_EXT: return TARGET_VECTOR; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; case ZVBB_EXT: return TARGET_ZVBB; case ZVBB_OR_ZVKB_EXT: diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..fd3ba29bae9 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlhu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlwu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vsb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vlsb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlshu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlswu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vssb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vlxb, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxh, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxw, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxbu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxhu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxwu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vsxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..0f3700d9269 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,168 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..3c12c124597 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..30bef369375 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..3c8b5ccc16b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..b7c00404f18 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..17a53012acf --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..b187cfc852b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-11 12:33 ` Re：[PATCH " juzhe.zhong @ 2024-01-11 12:36 ` joshua 0 siblings, 0 replies; 133+ messages in thread From: joshua @ 2024-01-11 12:36 UTC (permalink / raw) To: juzhe.zhong, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu Perhaps not. I suggest that is the limitation of rvv 0.7.1. I will consult my colleagues who are familiar with using rvv 0.7.1 tomorrow. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 20:33 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Does theadvector has extension instructions ? Show me the pattern. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 20:31 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Yes. (insn 14 13 15 2 (set (reg:QI 147) (const_int 1 [0x1])) {*movqi_internal} (nil)) ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 20:28 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. (vec_duplicate:RVVM1QI (reg:QI 147)))) Find the RTL define pseudo 147 to me. I guess it is 1. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 20:18 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. No, we have debugged and the code won't enter this autovec pattern. The root cause is CSE, in a very early pass. What I sent to you just now is the dump result which shows the difference. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 20:13 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. I guess it is because of patterns from autovectorization pattern: ;; Use define_insn_and_split to define vsext.vf2/vzext.vf2 will help ;; to combine instructions as below: ;; vsext.vf2 + vsext.vf2 + vadd.vv ==> vwadd.vv (define_insn_and_split "<optab><v_double_trunc><mode>2" [(set (match_operand:VWEXTI 0 "register_operand" "=&vr") (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 1 "register_operand" "vr")))] "TARGET_VECTOR && can_create_pseudo_p ()" "#" "&& 1" [(const_int 0)] { insn_code icode = code_for_pred_vf2 (<CODE>, <MODE>mode); riscv_vector::emit_vlmax_insn (icode, riscv_vector::UNARY_OP, operands); DONE; } [(set_attr "type" "vext") (set_attr "mode" "<MODE>")]) juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 20:05 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. without CSE: (insn 16 15 17 2 (set (reg/v:RVVM2HI 137 [ output_var_0 ]) (if_then_else:RVVM2HI (unspec:RVVMF8BI [ (const_vector:RVVMF8BI repeat [ (const_int 1 [0x1]) ]) (reg:DI 146) (const_int 2 [0x2]) repeated x2 (const_int 0 [0]) (reg:SI 66 vl) (reg:SI 67 vtype) ] UNSPEC_VPREDICATE) (mult:RVVM2HI (sign_extend:RVVM2HI (reg/v:RVVM1QI 136 [ op1 ])) (sign_extend:RVVM2HI (vec_duplicate:RVVM1QI (reg:QI 147)))) (unspec:RVVM2HI [ (reg:SI 0 zero) ] UNSPEC_VUNDEF))) {pred_dual_widen_mulsrvvm2hi_scalar} with CSE: (insn 16 15 17 2 (set (reg/v:RVVM2HI 137 [ output_var_0 ]) (if_then_else:RVVM2HI (unspec:RVVMF8BI [ (const_vector:RVVMF8BI repeat [ (const_int 1 [0x1]) ]) (reg:DI 146) (const_int 2 [0x2]) repeated x2 (const_int 0 [0]) (reg:SI 66 vl) (reg:SI 67 vtype) ] UNSPEC_VPREDICATE) (sign_extend:RVVM2HI (reg/v:RVVM1QI 136 [ op1 ])) (unspec:RVVM2HI [ (reg:SI 0 zero) ] UNSPEC_VUNDEF))) {pred_extendrvvm2hi_vf2} ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:32 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Ok. Let's hold on "RISC-V: Handle differences between XTheadvector and Vector" patch until you can reproduce the issue for me. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:29 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. The sext/zext issue is not related to xtheadvector-special patterns. I added !TARGET_XTHEADVECTOR to sext/zext patterns in "RISC-V: Handle differences between XTheadvector and Vector" That is caused by the vwmul pattern, but I cannot reproduce it right now. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:24 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. I don't see any patterns are possible CSE into sext/zext patterns: +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:21 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. "I didn't see theadvector-specific extension patterns. Could you show me?" They are all in the file thead-vector.md. For the sext/zext issue, perhaps I need some time to reproduce that optimization, but I can clearly remember it is related to vwmul. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:17 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. You mean which pattern optimized sext/vzext pattern? I didn't see theadvector-specific extension patterns. Could you show me? juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:14 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. To be specific, in CSE pass, the initial pattern will be optimized into the sext/zext pattern. ------------------------------------------------------------------ 发件人：joshua <cooper.joshua@linux.alibaba.com> 发送时间：2024年1月11日(星期四) 17:11 收件人："juzhe.zhong@rivai.ai"<juzhe.zhong@rivai.ai>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. sext/zext will be generated in O2 even without corresponding intrinsics. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:07 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ Add theadvector to the end of the enum. + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; Same. + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" You can use whole_reg_to_reg_move_p Btw, I review again : RISC-V: Handle differences between XTheadvector and Vector (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3713,7 +3744,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3734,7 +3765,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") Why do you add these !TARGERT_XTHEADVECRTOR ? juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-11 16:46 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (class th_extract): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (struct th_extract_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): * config/riscv/riscv-vector-builtins.h (enum required_ext): (struct function_group_info): * config/riscv/t-riscv: Add thead-vector-builtins-functions.def * config/riscv/thead-vector.md (@pred_mov_width<vlmem_op_attr><mode>): Add new patterns. (*pred_mov_width<vlmem_op_attr><mode>): Likewise. (@pred_store_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_load_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_store_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_load_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>): (@pred_th_extract<mode>): Likewise. (*pred_th_extract<mode>): Likewise. * config/riscv/thead-vector-builtins-functions.def: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 +++ .../riscv/riscv-vector-builtins-shapes.cc | 160 +++++++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + gcc/config/riscv/riscv-vector-builtins.cc | 70 +++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 253 ++++++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 +++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 +++++ 15 files changed, 1107 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 46f1a1da33e..3eba7943757 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2125,6 +2125,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2587,6 +2664,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> vsuxw_obj; +static CONSTEXPR const th_extract vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2878,6 +2986,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (vlb) +BASE (vlh) +BASE (vlw) +BASE (vlbu) +BASE (vlhu) +BASE (vlwu) +BASE (vsb) +BASE (vsh) +BASE (vsw) +BASE (vlsb) +BASE (vlsh) +BASE (vlsw) +BASE (vlsbu) +BASE (vlshu) +BASE (vlswu) +BASE (vssb) +BASE (vssh) +BASE (vssw) +BASE (vlxb) +BASE (vlxh) +BASE (vlxw) +BASE (vlxbu) +BASE (vlxhu) +BASE (vlxwu) +BASE (vsxb) +BASE (vsxh) +BASE (vsxw) +BASE (vsuxb) +BASE (vsuxh) +BASE (vsuxw) +BASE (vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 1122e3801a7..565a0311d2b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -299,6 +299,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const vlb; +extern const function_base *const vlh; +extern const function_base *const vlw; +extern const function_base *const vlbu; +extern const function_base *const vlhu; +extern const function_base *const vlwu; +extern const function_base *const vsb; +extern const function_base *const vsh; +extern const function_base *const vsw; +extern const function_base *const vlsb; +extern const function_base *const vlsh; +extern const function_base *const vlsw; +extern const function_base *const vlsbu; +extern const function_base *const vlshu; +extern const function_base *const vlswu; +extern const function_base *const vssb; +extern const function_base *const vssh; +extern const function_base *const vssw; +extern const function_base *const vlxb; +extern const function_base *const vlxh; +extern const function_base *const vlxw; +extern const function_base *const vlxbu; +extern const function_base *const vlxhu; +extern const function_base *const vlxwu; +extern const function_base *const vsxb; +extern const function_base *const vsxh; +extern const function_base *const vsxw; +extern const function_base *const vsuxb; +extern const function_base *const vsuxh; +extern const function_base *const vsuxw; +extern const function_base *const vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..8e90b17a94b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,146 @@ struct indexed_loadstore_def : public function_shape } }; +/* Add one function instance for GROUP, using operand suffix at index OI, + mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ +static void +build_th_loadstore (function_builder &b, const function_group_info &group, + unsigned int pred_idx, unsigned int vec_type_idx) +{ + auto_vec<tree, 5> argument_types; + function_instance function_instance (group.base_name, *group.base, + *group.shape, + group.ops_infos.types[vec_type_idx], + group.preds[pred_idx], &group.ops_infos); + tree return_type = group.ops_infos.ret.get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + b.allocate_argument_types (function_instance, argument_types); + b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + + tree type = builtin_types[group.ops_infos.types[vec_type_idx].index].vector; + if (strstr (group.base_name, "l") + && strstr (group.base_name, "u") + && !TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + if (strstr (group.base_name, "l") + && !strstr (group.base_name, "u") + && TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + machine_mode mode = TYPE_MODE (type); + int sew = GET_MODE_BITSIZE (GET_MODE_INNER (mode)); + if (strstr (group.base_name, "h") && sew == 8) + return; + + if (strstr (group.base_name, "w") && (sew == 8 || sew ==16)) + return; + + b.add_overloaded_function (function_instance, *group.shape); + b.add_unique_function (function_instance, (*group.shape), return_type, + argument_types); +} + +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +772,23 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1251,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1265,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..44b9fec1898 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -934,6 +934,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1481,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2672,38 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * index_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +2882,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..a8ee39a3cb2 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -114,6 +114,7 @@ static const unsigned int CP_WRITE_CSR = 1U << 5; enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ @@ -234,6 +235,8 @@ struct function_group_info { case VECTOR_EXT: return TARGET_VECTOR; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; case ZVBB_EXT: return TARGET_ZVBB; case ZVBB_OR_ZVKB_EXT: diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..fd3ba29bae9 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlhu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlwu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vsb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vlsb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlshu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlswu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vssb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vlxb, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxh, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxw, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxbu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxhu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxwu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vsxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..0f3700d9269 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,168 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..3c12c124597 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..30bef369375 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..3c8b5ccc16b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..b7c00404f18 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..17a53012acf --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..b187cfc852b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-11 9:17 ` juzhe.zhong 2024-01-11 9:21 ` Re：Re：[PATCH " joshua @ 2024-01-11 9:26 ` joshua 2024-01-11 9:28 ` Re：[PATCH " juzhe.zhong 1 sibling, 1 reply; 133+ messages in thread From: joshua @ 2024-01-11 9:26 UTC (permalink / raw) To: juzhe.zhong, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu Maybe the optimization cannot be done in simple cases. We run some complex cases in O2 and dsicovered it. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:17 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. You mean which pattern optimized sext/vzext pattern? I didn't see theadvector-specific extension patterns. Could you show me? juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:14 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. To be specific, in CSE pass, the initial pattern will be optimized into the sext/zext pattern. ------------------------------------------------------------------ 发件人：joshua <cooper.joshua@linux.alibaba.com> 发送时间：2024年1月11日(星期四) 17:11 收件人："juzhe.zhong@rivai.ai"<juzhe.zhong@rivai.ai>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. sext/zext will be generated in O2 even without corresponding intrinsics. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:07 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ Add theadvector to the end of the enum. + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; Same. + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" You can use whole_reg_to_reg_move_p Btw, I review again : RISC-V: Handle differences between XTheadvector and Vector (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3713,7 +3744,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3734,7 +3765,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") Why do you add these !TARGERT_XTHEADVECRTOR ? juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-11 16:46 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (class th_extract): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (struct th_extract_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): * config/riscv/riscv-vector-builtins.h (enum required_ext): (struct function_group_info): * config/riscv/t-riscv: Add thead-vector-builtins-functions.def * config/riscv/thead-vector.md (@pred_mov_width<vlmem_op_attr><mode>): Add new patterns. (*pred_mov_width<vlmem_op_attr><mode>): Likewise. (@pred_store_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_load_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_store_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_load_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>): (@pred_th_extract<mode>): Likewise. (*pred_th_extract<mode>): Likewise. * config/riscv/thead-vector-builtins-functions.def: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 +++ .../riscv/riscv-vector-builtins-shapes.cc | 160 +++++++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + gcc/config/riscv/riscv-vector-builtins.cc | 70 +++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 253 ++++++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 +++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 +++++ 15 files changed, 1107 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 46f1a1da33e..3eba7943757 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2125,6 +2125,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2587,6 +2664,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> vsuxw_obj; +static CONSTEXPR const th_extract vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2878,6 +2986,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (vlb) +BASE (vlh) +BASE (vlw) +BASE (vlbu) +BASE (vlhu) +BASE (vlwu) +BASE (vsb) +BASE (vsh) +BASE (vsw) +BASE (vlsb) +BASE (vlsh) +BASE (vlsw) +BASE (vlsbu) +BASE (vlshu) +BASE (vlswu) +BASE (vssb) +BASE (vssh) +BASE (vssw) +BASE (vlxb) +BASE (vlxh) +BASE (vlxw) +BASE (vlxbu) +BASE (vlxhu) +BASE (vlxwu) +BASE (vsxb) +BASE (vsxh) +BASE (vsxw) +BASE (vsuxb) +BASE (vsuxh) +BASE (vsuxw) +BASE (vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 1122e3801a7..565a0311d2b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -299,6 +299,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const vlb; +extern const function_base *const vlh; +extern const function_base *const vlw; +extern const function_base *const vlbu; +extern const function_base *const vlhu; +extern const function_base *const vlwu; +extern const function_base *const vsb; +extern const function_base *const vsh; +extern const function_base *const vsw; +extern const function_base *const vlsb; +extern const function_base *const vlsh; +extern const function_base *const vlsw; +extern const function_base *const vlsbu; +extern const function_base *const vlshu; +extern const function_base *const vlswu; +extern const function_base *const vssb; +extern const function_base *const vssh; +extern const function_base *const vssw; +extern const function_base *const vlxb; +extern const function_base *const vlxh; +extern const function_base *const vlxw; +extern const function_base *const vlxbu; +extern const function_base *const vlxhu; +extern const function_base *const vlxwu; +extern const function_base *const vsxb; +extern const function_base *const vsxh; +extern const function_base *const vsxw; +extern const function_base *const vsuxb; +extern const function_base *const vsuxh; +extern const function_base *const vsuxw; +extern const function_base *const vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..8e90b17a94b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,146 @@ struct indexed_loadstore_def : public function_shape } }; +/* Add one function instance for GROUP, using operand suffix at index OI, + mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ +static void +build_th_loadstore (function_builder &b, const function_group_info &group, + unsigned int pred_idx, unsigned int vec_type_idx) +{ + auto_vec<tree, 5> argument_types; + function_instance function_instance (group.base_name, *group.base, + *group.shape, + group.ops_infos.types[vec_type_idx], + group.preds[pred_idx], &group.ops_infos); + tree return_type = group.ops_infos.ret.get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + b.allocate_argument_types (function_instance, argument_types); + b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + + tree type = builtin_types[group.ops_infos.types[vec_type_idx].index].vector; + if (strstr (group.base_name, "l") + && strstr (group.base_name, "u") + && !TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + if (strstr (group.base_name, "l") + && !strstr (group.base_name, "u") + && TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + machine_mode mode = TYPE_MODE (type); + int sew = GET_MODE_BITSIZE (GET_MODE_INNER (mode)); + if (strstr (group.base_name, "h") && sew == 8) + return; + + if (strstr (group.base_name, "w") && (sew == 8 || sew ==16)) + return; + + b.add_overloaded_function (function_instance, *group.shape); + b.add_unique_function (function_instance, (*group.shape), return_type, + argument_types); +} + +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +772,23 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1251,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1265,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..44b9fec1898 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -934,6 +934,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1481,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2672,38 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * index_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +2882,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..a8ee39a3cb2 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -114,6 +114,7 @@ static const unsigned int CP_WRITE_CSR = 1U << 5; enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ @@ -234,6 +235,8 @@ struct function_group_info { case VECTOR_EXT: return TARGET_VECTOR; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; case ZVBB_EXT: return TARGET_ZVBB; case ZVBB_OR_ZVKB_EXT: diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..fd3ba29bae9 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlhu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlwu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vsb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vlsb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlshu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlswu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vssb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vlxb, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxh, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxw, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxbu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxhu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxwu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vsxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..0f3700d9269 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,168 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..3c12c124597 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..30bef369375 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..3c8b5ccc16b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..b7c00404f18 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..17a53012acf --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..b187cfc852b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-11 9:26 ` joshua @ 2024-01-11 9:28 ` juzhe.zhong 2024-01-11 9:35 ` Re：Re：[PATCH " joshua 0 siblings, 1 reply; 133+ messages in thread From: juzhe.zhong @ 2024-01-11 9:28 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 60213 bytes --] I prefer you remove those TARGET_THEADVECTOR for now. And file PR let me see the real problem. I don't believe this should not fixed by this way. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:26 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Maybe the optimization cannot be done in simple cases. We run some complex cases in O2 and dsicovered it. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:17 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. You mean which pattern optimized sext/vzext pattern? I didn't see theadvector-specific extension patterns. Could you show me? juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:14 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. To be specific, in CSE pass, the initial pattern will be optimized into the sext/zext pattern. ------------------------------------------------------------------ 发件人：joshua <cooper.joshua@linux.alibaba.com> 发送时间：2024年1月11日(星期四) 17:11 收件人："juzhe.zhong@rivai.ai"<juzhe.zhong@rivai.ai>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. sext/zext will be generated in O2 even without corresponding intrinsics. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:07 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ Add theadvector to the end of the enum. + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; Same. + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" You can use whole_reg_to_reg_move_p Btw, I review again : RISC-V: Handle differences between XTheadvector and Vector (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3713,7 +3744,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3734,7 +3765,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") Why do you add these !TARGERT_XTHEADVECRTOR ? juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-11 16:46 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (class th_extract): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (struct th_extract_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): * config/riscv/riscv-vector-builtins.h (enum required_ext): (struct function_group_info): * config/riscv/t-riscv: Add thead-vector-builtins-functions.def * config/riscv/thead-vector.md (@pred_mov_width<vlmem_op_attr><mode>): Add new patterns. (*pred_mov_width<vlmem_op_attr><mode>): Likewise. (@pred_store_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_load_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_store_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_load_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>): (@pred_th_extract<mode>): Likewise. (*pred_th_extract<mode>): Likewise. * config/riscv/thead-vector-builtins-functions.def: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 +++ .../riscv/riscv-vector-builtins-shapes.cc | 160 +++++++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + gcc/config/riscv/riscv-vector-builtins.cc | 70 +++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 253 ++++++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 +++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 +++++ 15 files changed, 1107 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 46f1a1da33e..3eba7943757 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2125,6 +2125,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2587,6 +2664,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> vsuxw_obj; +static CONSTEXPR const th_extract vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2878,6 +2986,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (vlb) +BASE (vlh) +BASE (vlw) +BASE (vlbu) +BASE (vlhu) +BASE (vlwu) +BASE (vsb) +BASE (vsh) +BASE (vsw) +BASE (vlsb) +BASE (vlsh) +BASE (vlsw) +BASE (vlsbu) +BASE (vlshu) +BASE (vlswu) +BASE (vssb) +BASE (vssh) +BASE (vssw) +BASE (vlxb) +BASE (vlxh) +BASE (vlxw) +BASE (vlxbu) +BASE (vlxhu) +BASE (vlxwu) +BASE (vsxb) +BASE (vsxh) +BASE (vsxw) +BASE (vsuxb) +BASE (vsuxh) +BASE (vsuxw) +BASE (vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 1122e3801a7..565a0311d2b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -299,6 +299,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const vlb; +extern const function_base *const vlh; +extern const function_base *const vlw; +extern const function_base *const vlbu; +extern const function_base *const vlhu; +extern const function_base *const vlwu; +extern const function_base *const vsb; +extern const function_base *const vsh; +extern const function_base *const vsw; +extern const function_base *const vlsb; +extern const function_base *const vlsh; +extern const function_base *const vlsw; +extern const function_base *const vlsbu; +extern const function_base *const vlshu; +extern const function_base *const vlswu; +extern const function_base *const vssb; +extern const function_base *const vssh; +extern const function_base *const vssw; +extern const function_base *const vlxb; +extern const function_base *const vlxh; +extern const function_base *const vlxw; +extern const function_base *const vlxbu; +extern const function_base *const vlxhu; +extern const function_base *const vlxwu; +extern const function_base *const vsxb; +extern const function_base *const vsxh; +extern const function_base *const vsxw; +extern const function_base *const vsuxb; +extern const function_base *const vsuxh; +extern const function_base *const vsuxw; +extern const function_base *const vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..8e90b17a94b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,146 @@ struct indexed_loadstore_def : public function_shape } }; +/* Add one function instance for GROUP, using operand suffix at index OI, + mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ +static void +build_th_loadstore (function_builder &b, const function_group_info &group, + unsigned int pred_idx, unsigned int vec_type_idx) +{ + auto_vec<tree, 5> argument_types; + function_instance function_instance (group.base_name, *group.base, + *group.shape, + group.ops_infos.types[vec_type_idx], + group.preds[pred_idx], &group.ops_infos); + tree return_type = group.ops_infos.ret.get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + b.allocate_argument_types (function_instance, argument_types); + b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + + tree type = builtin_types[group.ops_infos.types[vec_type_idx].index].vector; + if (strstr (group.base_name, "l") + && strstr (group.base_name, "u") + && !TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + if (strstr (group.base_name, "l") + && !strstr (group.base_name, "u") + && TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + machine_mode mode = TYPE_MODE (type); + int sew = GET_MODE_BITSIZE (GET_MODE_INNER (mode)); + if (strstr (group.base_name, "h") && sew == 8) + return; + + if (strstr (group.base_name, "w") && (sew == 8 || sew ==16)) + return; + + b.add_overloaded_function (function_instance, *group.shape); + b.add_unique_function (function_instance, (*group.shape), return_type, + argument_types); +} + +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +772,23 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1251,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1265,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..44b9fec1898 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -934,6 +934,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1481,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2672,38 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * index_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +2882,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..a8ee39a3cb2 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -114,6 +114,7 @@ static const unsigned int CP_WRITE_CSR = 1U << 5; enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ @@ -234,6 +235,8 @@ struct function_group_info { case VECTOR_EXT: return TARGET_VECTOR; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; case ZVBB_EXT: return TARGET_ZVBB; case ZVBB_OR_ZVKB_EXT: diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..fd3ba29bae9 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlhu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlwu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vsb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vlsb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlshu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlswu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vssb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vlxb, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxh, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxw, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxbu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxhu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxwu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vsxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..0f3700d9269 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,168 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..3c12c124597 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..30bef369375 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..3c8b5ccc16b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..b7c00404f18 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..17a53012acf --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..b187cfc852b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-11 9:28 ` Re：[PATCH " juzhe.zhong @ 2024-01-11 9:35 ` joshua 0 siblings, 0 replies; 133+ messages in thread From: joshua @ 2024-01-11 9:35 UTC (permalink / raw) To: juzhe.zhong, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu I see. We added ! TARGET_THEADVECTOR initally becasue we want to provide a patchset version that can work without any errors both in O0 and O2. Without these changes, we will get " unrecognized opcode" in O2 during assembly stage. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:28 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. I prefer you remove those TARGET_THEADVECTOR for now. And file PR let me see the real problem. I don't believe this should not fixed by this way. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:26 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Maybe the optimization cannot be done in simple cases. We run some complex cases in O2 and dsicovered it. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:17 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. You mean which pattern optimized sext/vzext pattern? I didn't see theadvector-specific extension patterns. Could you show me? juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 17:14 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. To be specific, in CSE pass, the initial pattern will be optimized into the sext/zext pattern. ------------------------------------------------------------------ 发件人：joshua <cooper.joshua@linux.alibaba.com> 发送时间：2024年1月11日(星期四) 17:11 收件人："juzhe.zhong@rivai.ai"<juzhe.zhong@rivai.ai>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. sext/zext will be generated in O2 even without corresponding intrinsics. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:07 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ Add theadvector to the end of the enum. + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; Same. + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" You can use whole_reg_to_reg_move_p Btw, I review again : RISC-V: Handle differences between XTheadvector and Vector (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3713,7 +3744,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>")@@ -3734,7 +3765,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))]- "TARGET_VECTOR"+ "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") Why do you add these !TARGERT_XTHEADVECRTOR ? juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-11 16:46 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (class th_extract): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (struct th_extract_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): * config/riscv/riscv-vector-builtins.h (enum required_ext): (struct function_group_info): * config/riscv/t-riscv: Add thead-vector-builtins-functions.def * config/riscv/thead-vector.md (@pred_mov_width<vlmem_op_attr><mode>): Add new patterns. (*pred_mov_width<vlmem_op_attr><mode>): Likewise. (@pred_store_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_load_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_store_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_load_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>): (@pred_th_extract<mode>): Likewise. (*pred_th_extract<mode>): Likewise. * config/riscv/thead-vector-builtins-functions.def: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 +++ .../riscv/riscv-vector-builtins-shapes.cc | 160 +++++++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + gcc/config/riscv/riscv-vector-builtins.cc | 70 +++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 253 ++++++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 +++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 +++++ 15 files changed, 1107 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 46f1a1da33e..3eba7943757 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2125,6 +2125,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2587,6 +2664,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> vsuxw_obj; +static CONSTEXPR const th_extract vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2878,6 +2986,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (vlb) +BASE (vlh) +BASE (vlw) +BASE (vlbu) +BASE (vlhu) +BASE (vlwu) +BASE (vsb) +BASE (vsh) +BASE (vsw) +BASE (vlsb) +BASE (vlsh) +BASE (vlsw) +BASE (vlsbu) +BASE (vlshu) +BASE (vlswu) +BASE (vssb) +BASE (vssh) +BASE (vssw) +BASE (vlxb) +BASE (vlxh) +BASE (vlxw) +BASE (vlxbu) +BASE (vlxhu) +BASE (vlxwu) +BASE (vsxb) +BASE (vsxh) +BASE (vsxw) +BASE (vsuxb) +BASE (vsuxh) +BASE (vsuxw) +BASE (vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 1122e3801a7..565a0311d2b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -299,6 +299,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const vlb; +extern const function_base *const vlh; +extern const function_base *const vlw; +extern const function_base *const vlbu; +extern const function_base *const vlhu; +extern const function_base *const vlwu; +extern const function_base *const vsb; +extern const function_base *const vsh; +extern const function_base *const vsw; +extern const function_base *const vlsb; +extern const function_base *const vlsh; +extern const function_base *const vlsw; +extern const function_base *const vlsbu; +extern const function_base *const vlshu; +extern const function_base *const vlswu; +extern const function_base *const vssb; +extern const function_base *const vssh; +extern const function_base *const vssw; +extern const function_base *const vlxb; +extern const function_base *const vlxh; +extern const function_base *const vlxw; +extern const function_base *const vlxbu; +extern const function_base *const vlxhu; +extern const function_base *const vlxwu; +extern const function_base *const vsxb; +extern const function_base *const vsxh; +extern const function_base *const vsxw; +extern const function_base *const vsuxb; +extern const function_base *const vsuxh; +extern const function_base *const vsuxw; +extern const function_base *const vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..8e90b17a94b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,146 @@ struct indexed_loadstore_def : public function_shape } }; +/* Add one function instance for GROUP, using operand suffix at index OI, + mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ +static void +build_th_loadstore (function_builder &b, const function_group_info &group, + unsigned int pred_idx, unsigned int vec_type_idx) +{ + auto_vec<tree, 5> argument_types; + function_instance function_instance (group.base_name, *group.base, + *group.shape, + group.ops_infos.types[vec_type_idx], + group.preds[pred_idx], &group.ops_infos); + tree return_type = group.ops_infos.ret.get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + b.allocate_argument_types (function_instance, argument_types); + b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + + tree type = builtin_types[group.ops_infos.types[vec_type_idx].index].vector; + if (strstr (group.base_name, "l") + && strstr (group.base_name, "u") + && !TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + if (strstr (group.base_name, "l") + && !strstr (group.base_name, "u") + && TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + machine_mode mode = TYPE_MODE (type); + int sew = GET_MODE_BITSIZE (GET_MODE_INNER (mode)); + if (strstr (group.base_name, "h") && sew == 8) + return; + + if (strstr (group.base_name, "w") && (sew == 8 || sew ==16)) + return; + + b.add_overloaded_function (function_instance, *group.shape); + b.add_unique_function (function_instance, (*group.shape), return_type, + argument_types); +} + +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +772,23 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1251,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1265,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..44b9fec1898 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -934,6 +934,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1481,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2672,38 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * index_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +2882,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..a8ee39a3cb2 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -114,6 +114,7 @@ static const unsigned int CP_WRITE_CSR = 1U << 5; enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ @@ -234,6 +235,8 @@ struct function_group_info { case VECTOR_EXT: return TARGET_VECTOR; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; case ZVBB_EXT: return TARGET_ZVBB; case ZVBB_OR_ZVKB_EXT: diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..fd3ba29bae9 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlhu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlwu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vsb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vlsb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlshu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlswu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vssb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vlxb, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxh, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxw, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxbu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxhu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxwu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vsxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..0f3700d9269 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,168 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..3c12c124597 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..30bef369375 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..3c8b5ccc16b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..b7c00404f18 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..17a53012acf --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..b187cfc852b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-11 9:07 ` juzhe.zhong 2024-01-11 9:11 ` Re：[PATCH " joshua @ 2024-01-11 9:54 ` joshua 1 sibling, 0 replies; 133+ messages in thread From: joshua @ 2024-01-11 9:54 UTC (permalink / raw) To: juzhe.zhong, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 54637 bytes --] The "Add support for xtheadvector-specific intrinsics" patch has been updated according to the comments. [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. (gnu.org) <https://gcc.gnu.org/pipermail/gcc-patches/2024-January/642627.html > ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:07 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ Add theadvector to the end of the enum. + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; Same. + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" You can use whole_reg_to_reg_move_p Btw, I review again : RISC-V: Handle differences between XTheadvector and Vector (any_extend:VWEXTI (match_operand:<V_DOUBLE_TRUNC> 3 "register_operand" "W21,W21,W21,W21,W42,W42,W42,W42,W84,W84,W84,W84, vr, vr")) (match_operand:VWEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf2\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3713,7 +3744,7 @@ (any_extend:VQEXTI (match_operand:<V_QUAD_TRUNC> 3 "register_operand" "W43,W43,W43,W43,W86,W86,W86,W86, vr, vr")) (match_operand:VQEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf4\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") @@ -3734,7 +3765,7 @@ (any_extend:VOEXTI (match_operand:<V_OCT_TRUNC> 3 "register_operand" "W87,W87,W87,W87, vr, vr")) (match_operand:VOEXTI 2 "vector_merge_operand" " vu, vu, 0, 0, vu, 0")))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "v<sz>ext.vf8\t%0,%3%p1" [(set_attr "type" "vext") (set_attr "mode" "<MODE>") Why do you add these !TARGERT_XTHEADVECRTOR ? juzhe.zhong@rivai.ai From: Jun Sha (Joshua) <mailto:cooper.joshua@linux.alibaba.com > Date: 2024-01-11 16:46 To: gcc-patches <mailto:gcc-patches@gcc.gnu.org > CC: jim.wilson.gcc <mailto:jim.wilson.gcc@gmail.com >; palmer <mailto:palmer@dabbelt.com >; andrew <mailto:andrew@sifive.com >; philipp.tomsich <mailto:philipp.tomsich@vrull.eu >; jeffreyalaw <mailto:jeffreyalaw@gmail.com >; christoph.muellner <mailto:christoph.muellner@vrull.eu >; juzhe.zhong <mailto:juzhe.zhong@rivai.ai >; Jun Sha (Joshua) <mailto:cooper.joshua@linux.alibaba.com >; Jin Ma <mailto:jinma@linux.alibaba.com >; Xianmiao Qu <mailto:cooper.qu@linux.alibaba.com > Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (class th_extract): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (struct th_extract_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): * config/riscv/riscv-vector-builtins.h (enum required_ext): (struct function_group_info): * config/riscv/t-riscv: Add thead-vector-builtins-functions.def * config/riscv/thead-vector.md (@pred_mov_width<vlmem_op_attr><mode>): Add new patterns. (*pred_mov_width<vlmem_op_attr><mode>): Likewise. (@pred_store_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_load_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_store_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_load_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>): (@pred_th_extract<mode>): Likewise. (*pred_th_extract<mode>): Likewise. * config/riscv/thead-vector-builtins-functions.def: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 +++ .../riscv/riscv-vector-builtins-shapes.cc | 160 +++++++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + gcc/config/riscv/riscv-vector-builtins.cc | 70 +++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 253 ++++++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 +++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 +++++ 15 files changed, 1107 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 46f1a1da33e..3eba7943757 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2125,6 +2125,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2587,6 +2664,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> vsuxw_obj; +static CONSTEXPR const th_extract vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2878,6 +2986,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (vlb) +BASE (vlh) +BASE (vlw) +BASE (vlbu) +BASE (vlhu) +BASE (vlwu) +BASE (vsb) +BASE (vsh) +BASE (vsw) +BASE (vlsb) +BASE (vlsh) +BASE (vlsw) +BASE (vlsbu) +BASE (vlshu) +BASE (vlswu) +BASE (vssb) +BASE (vssh) +BASE (vssw) +BASE (vlxb) +BASE (vlxh) +BASE (vlxw) +BASE (vlxbu) +BASE (vlxhu) +BASE (vlxwu) +BASE (vsxb) +BASE (vsxh) +BASE (vsxw) +BASE (vsuxb) +BASE (vsuxh) +BASE (vsuxw) +BASE (vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 1122e3801a7..565a0311d2b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -299,6 +299,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const vlb; +extern const function_base *const vlh; +extern const function_base *const vlw; +extern const function_base *const vlbu; +extern const function_base *const vlhu; +extern const function_base *const vlwu; +extern const function_base *const vsb; +extern const function_base *const vsh; +extern const function_base *const vsw; +extern const function_base *const vlsb; +extern const function_base *const vlsh; +extern const function_base *const vlsw; +extern const function_base *const vlsbu; +extern const function_base *const vlshu; +extern const function_base *const vlswu; +extern const function_base *const vssb; +extern const function_base *const vssh; +extern const function_base *const vssw; +extern const function_base *const vlxb; +extern const function_base *const vlxh; +extern const function_base *const vlxw; +extern const function_base *const vlxbu; +extern const function_base *const vlxhu; +extern const function_base *const vlxwu; +extern const function_base *const vsxb; +extern const function_base *const vsxh; +extern const function_base *const vsxw; +extern const function_base *const vsuxb; +extern const function_base *const vsuxh; +extern const function_base *const vsuxw; +extern const function_base *const vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..8e90b17a94b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,146 @@ struct indexed_loadstore_def : public function_shape } }; +/* Add one function instance for GROUP, using operand suffix at index OI, + mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ +static void +build_th_loadstore (function_builder &b, const function_group_info &group, + unsigned int pred_idx, unsigned int vec_type_idx) +{ + auto_vec<tree, 5> argument_types; + function_instance function_instance (group.base_name, *group.base, + *group.shape, + group.ops_infos.types[vec_type_idx], + group.preds[pred_idx], &group.ops_infos); + tree return_type = group.ops_infos.ret.get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + b.allocate_argument_types (function_instance, argument_types); + b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + + tree type = builtin_types[group.ops_infos.types[vec_type_idx].index].vector; + if (strstr (group.base_name, "l") + && strstr (group.base_name, "u") + && !TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + if (strstr (group.base_name, "l") + && !strstr (group.base_name, "u") + && TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + machine_mode mode = TYPE_MODE (type); + int sew = GET_MODE_BITSIZE (GET_MODE_INNER (mode)); + if (strstr (group.base_name, "h") && sew == 8) + return; + + if (strstr (group.base_name, "w") && (sew == 8 || sew ==16)) + return; + + b.add_overloaded_function (function_instance, *group.shape); + b.add_unique_function (function_instance, (*group.shape), return_type, + argument_types); +} + +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +772,23 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1251,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1265,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..44b9fec1898 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -934,6 +934,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1481,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2672,38 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * index_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +2882,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..a8ee39a3cb2 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -114,6 +114,7 @@ static const unsigned int CP_WRITE_CSR = 1U << 5; enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ @@ -234,6 +235,8 @@ struct function_group_info { case VECTOR_EXT: return TARGET_VECTOR; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; case ZVBB_EXT: return TARGET_ZVBB; case ZVBB_OR_ZVKB_EXT: diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..fd3ba29bae9 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlhu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlwu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vsb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vlsb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlshu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlswu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vssb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vlxb, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxh, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxw, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxbu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxhu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxwu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vsxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..0f3700d9269 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,168 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..3c12c124597 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..30bef369375 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..3c8b5ccc16b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..b7c00404f18 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..17a53012acf --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..b187cfc852b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-10 9:31 ` [PATCH " Jun Sha (Joshua) 2024-01-11 8:46 ` Jun Sha (Joshua) @ 2024-01-11 9:52 ` Jun Sha (Joshua) 2024-01-11 9:57 ` juzhe.zhong 1 sibling, 1 reply; 133+ messages in thread From: Jun Sha (Joshua) @ 2024-01-11 9:52 UTC (permalink / raw) To: gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, Jun Sha (Joshua), Jin Ma, Xianmiao Qu This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (class th_extract): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (struct th_extract_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): * config/riscv/riscv-vector-builtins.h (enum required_ext): (struct function_group_info): * config/riscv/t-riscv: Add thead-vector-builtins-functions.def * config/riscv/thead-vector.md (@pred_mov_width<vlmem_op_attr><mode>): Add new patterns. (*pred_mov_width<vlmem_op_attr><mode>): Likewise. (@pred_store_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_load_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_store_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_load_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>): (@pred_th_extract<mode>): Likewise. (*pred_th_extract<mode>): Likewise. * config/riscv/thead-vector-builtins-functions.def: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 +++ .../riscv/riscv-vector-builtins-shapes.cc | 160 +++++++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + gcc/config/riscv/riscv-vector-builtins.cc | 70 +++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 250 ++++++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 +++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 +++++ 15 files changed, 1104 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 1aa6e3c6665..b6f6e4ff37e 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2141,6 +2141,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2603,6 +2680,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> vsuxw_obj; +static CONSTEXPR const th_extract vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2894,6 +3002,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (vlb) +BASE (vlh) +BASE (vlw) +BASE (vlbu) +BASE (vlhu) +BASE (vlwu) +BASE (vsb) +BASE (vsh) +BASE (vsw) +BASE (vlsb) +BASE (vlsh) +BASE (vlsw) +BASE (vlsbu) +BASE (vlshu) +BASE (vlswu) +BASE (vssb) +BASE (vssh) +BASE (vssw) +BASE (vlxb) +BASE (vlxh) +BASE (vlxw) +BASE (vlxbu) +BASE (vlxhu) +BASE (vlxwu) +BASE (vsxb) +BASE (vsxh) +BASE (vsxw) +BASE (vsuxb) +BASE (vsuxh) +BASE (vsuxw) +BASE (vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 87c7f43a423..1f2c94d3541 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -280,6 +280,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const vlb; +extern const function_base *const vlh; +extern const function_base *const vlw; +extern const function_base *const vlbu; +extern const function_base *const vlhu; +extern const function_base *const vlwu; +extern const function_base *const vsb; +extern const function_base *const vsh; +extern const function_base *const vsw; +extern const function_base *const vlsb; +extern const function_base *const vlsh; +extern const function_base *const vlsw; +extern const function_base *const vlsbu; +extern const function_base *const vlshu; +extern const function_base *const vlswu; +extern const function_base *const vssb; +extern const function_base *const vssh; +extern const function_base *const vssw; +extern const function_base *const vlxb; +extern const function_base *const vlxh; +extern const function_base *const vlxw; +extern const function_base *const vlxbu; +extern const function_base *const vlxhu; +extern const function_base *const vlxwu; +extern const function_base *const vsxb; +extern const function_base *const vsxh; +extern const function_base *const vsxw; +extern const function_base *const vsuxb; +extern const function_base *const vsuxh; +extern const function_base *const vsuxw; +extern const function_base *const vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..8e90b17a94b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,146 @@ struct indexed_loadstore_def : public function_shape } }; +/* Add one function instance for GROUP, using operand suffix at index OI, + mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ +static void +build_th_loadstore (function_builder &b, const function_group_info &group, + unsigned int pred_idx, unsigned int vec_type_idx) +{ + auto_vec<tree, 5> argument_types; + function_instance function_instance (group.base_name, *group.base, + *group.shape, + group.ops_infos.types[vec_type_idx], + group.preds[pred_idx], &group.ops_infos); + tree return_type = group.ops_infos.ret.get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + b.allocate_argument_types (function_instance, argument_types); + b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + + tree type = builtin_types[group.ops_infos.types[vec_type_idx].index].vector; + if (strstr (group.base_name, "l") + && strstr (group.base_name, "u") + && !TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + if (strstr (group.base_name, "l") + && !strstr (group.base_name, "u") + && TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + machine_mode mode = TYPE_MODE (type); + int sew = GET_MODE_BITSIZE (GET_MODE_INNER (mode)); + if (strstr (group.base_name, "h") && sew == 8) + return; + + if (strstr (group.base_name, "w") && (sew == 8 || sew ==16)) + return; + + b.add_overloaded_function (function_instance, *group.shape); + b.add_unique_function (function_instance, (*group.shape), return_type, + argument_types); +} + +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +772,23 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1251,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1265,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..44b9fec1898 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -934,6 +934,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1481,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2672,38 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * index_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +2882,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..22fed60b4c3 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -123,6 +123,7 @@ enum required_ext ZVKNHB_EXT, /* Crypto vector Zvknhb sub-ext */ ZVKSED_EXT, /* Crypto vector Zvksed sub-ext */ ZVKSH_EXT, /* Crypto vector Zvksh sub-ext */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ }; /* Enumerates the RVV operand types. */ @@ -252,6 +253,8 @@ struct function_group_info return TARGET_ZVKSED; case ZVKSH_EXT: return TARGET_ZVKSH; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; default: gcc_unreachable (); } diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..fd3ba29bae9 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlhu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlwu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vsb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vlsb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlshu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlswu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vssb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vlxb, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxh, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxw, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxbu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxhu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxwu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vsxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..5fe9ba08c4e 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,165 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& riscv_vector::whole_reg_to_reg_move_p (operands, <MODE>mode, 7)" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..3c12c124597 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..30bef369375 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..3c8b5ccc16b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..b7c00404f18 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..17a53012acf --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..b187cfc852b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-11 9:52 ` [PATCH " Jun Sha (Joshua) @ 2024-01-11 9:57 ` juzhe.zhong 2024-01-11 10:54 ` Re：[PATCH " joshua 0 siblings, 1 reply; 133+ messages in thread From: juzhe.zhong @ 2024-01-11 9:57 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, cooper.joshua, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 54711 bytes --] LGTM. Could you resend the patch "RISC-V: Handle differences between XTheadvector and Vector Thanks. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-11 17:52 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (class th_extract): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (struct th_extract_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): * config/riscv/riscv-vector-builtins.h (enum required_ext): (struct function_group_info): * config/riscv/t-riscv: Add thead-vector-builtins-functions.def * config/riscv/thead-vector.md (@pred_mov_width<vlmem_op_attr><mode>): Add new patterns. (*pred_mov_width<vlmem_op_attr><mode>): Likewise. (@pred_store_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_load_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_store_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_load_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>): (@pred_th_extract<mode>): Likewise. (*pred_th_extract<mode>): Likewise. * config/riscv/thead-vector-builtins-functions.def: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 +++ .../riscv/riscv-vector-builtins-shapes.cc | 160 +++++++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + gcc/config/riscv/riscv-vector-builtins.cc | 70 +++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 250 ++++++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 +++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 +++++ 15 files changed, 1104 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 1aa6e3c6665..b6f6e4ff37e 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2141,6 +2141,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2603,6 +2680,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> vsuxw_obj; +static CONSTEXPR const th_extract vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2894,6 +3002,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (vlb) +BASE (vlh) +BASE (vlw) +BASE (vlbu) +BASE (vlhu) +BASE (vlwu) +BASE (vsb) +BASE (vsh) +BASE (vsw) +BASE (vlsb) +BASE (vlsh) +BASE (vlsw) +BASE (vlsbu) +BASE (vlshu) +BASE (vlswu) +BASE (vssb) +BASE (vssh) +BASE (vssw) +BASE (vlxb) +BASE (vlxh) +BASE (vlxw) +BASE (vlxbu) +BASE (vlxhu) +BASE (vlxwu) +BASE (vsxb) +BASE (vsxh) +BASE (vsxw) +BASE (vsuxb) +BASE (vsuxh) +BASE (vsuxw) +BASE (vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 87c7f43a423..1f2c94d3541 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -280,6 +280,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const vlb; +extern const function_base *const vlh; +extern const function_base *const vlw; +extern const function_base *const vlbu; +extern const function_base *const vlhu; +extern const function_base *const vlwu; +extern const function_base *const vsb; +extern const function_base *const vsh; +extern const function_base *const vsw; +extern const function_base *const vlsb; +extern const function_base *const vlsh; +extern const function_base *const vlsw; +extern const function_base *const vlsbu; +extern const function_base *const vlshu; +extern const function_base *const vlswu; +extern const function_base *const vssb; +extern const function_base *const vssh; +extern const function_base *const vssw; +extern const function_base *const vlxb; +extern const function_base *const vlxh; +extern const function_base *const vlxw; +extern const function_base *const vlxbu; +extern const function_base *const vlxhu; +extern const function_base *const vlxwu; +extern const function_base *const vsxb; +extern const function_base *const vsxh; +extern const function_base *const vsxw; +extern const function_base *const vsuxb; +extern const function_base *const vsuxh; +extern const function_base *const vsuxw; +extern const function_base *const vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..8e90b17a94b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,146 @@ struct indexed_loadstore_def : public function_shape } }; +/* Add one function instance for GROUP, using operand suffix at index OI, + mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ +static void +build_th_loadstore (function_builder &b, const function_group_info &group, + unsigned int pred_idx, unsigned int vec_type_idx) +{ + auto_vec<tree, 5> argument_types; + function_instance function_instance (group.base_name, *group.base, + *group.shape, + group.ops_infos.types[vec_type_idx], + group.preds[pred_idx], &group.ops_infos); + tree return_type = group.ops_infos.ret.get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + b.allocate_argument_types (function_instance, argument_types); + b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + + tree type = builtin_types[group.ops_infos.types[vec_type_idx].index].vector; + if (strstr (group.base_name, "l") + && strstr (group.base_name, "u") + && !TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + if (strstr (group.base_name, "l") + && !strstr (group.base_name, "u") + && TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + machine_mode mode = TYPE_MODE (type); + int sew = GET_MODE_BITSIZE (GET_MODE_INNER (mode)); + if (strstr (group.base_name, "h") && sew == 8) + return; + + if (strstr (group.base_name, "w") && (sew == 8 || sew ==16)) + return; + + b.add_overloaded_function (function_instance, *group.shape); + b.add_unique_function (function_instance, (*group.shape), return_type, + argument_types); +} + +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +772,23 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1251,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1265,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..44b9fec1898 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -934,6 +934,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1481,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2672,38 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * index_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +2882,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..22fed60b4c3 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -123,6 +123,7 @@ enum required_ext ZVKNHB_EXT, /* Crypto vector Zvknhb sub-ext */ ZVKSED_EXT, /* Crypto vector Zvksed sub-ext */ ZVKSH_EXT, /* Crypto vector Zvksh sub-ext */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ }; /* Enumerates the RVV operand types. */ @@ -252,6 +253,8 @@ struct function_group_info return TARGET_ZVKSED; case ZVKSH_EXT: return TARGET_ZVKSH; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; default: gcc_unreachable (); } diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..fd3ba29bae9 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlhu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlwu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vsb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vlsb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlshu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlswu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vssb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vlxb, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxh, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxw, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxbu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxhu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxwu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vsxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..5fe9ba08c4e 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,165 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& riscv_vector::whole_reg_to_reg_move_p (operands, <MODE>mode, 7)" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..3c12c124597 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..30bef369375 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..3c8b5ccc16b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..b7c00404f18 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..17a53012acf --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..b187cfc852b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-11 9:57 ` juzhe.zhong @ 2024-01-11 10:54 ` joshua 2024-01-11 10:55 ` juzhe.zhong 0 siblings, 1 reply; 133+ messages in thread From: joshua @ 2024-01-11 10:54 UTC (permalink / raw) To: juzhe.zhong, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu Do you mean removing TARGET_XTHEADVECTOR for sext/zext patterns and then resending the "RISC-V: Handle differences between XTheadvector and Vector" patch? ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:57 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. LGTM. Could you resend the patch "RISC-V: Handle differences between XTheadvector and Vector Thanks. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-11 17:52 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (class th_extract): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (struct th_extract_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): * config/riscv/riscv-vector-builtins.h (enum required_ext): (struct function_group_info): * config/riscv/t-riscv: Add thead-vector-builtins-functions.def * config/riscv/thead-vector.md (@pred_mov_width<vlmem_op_attr><mode>): Add new patterns. (*pred_mov_width<vlmem_op_attr><mode>): Likewise. (@pred_store_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_load_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_store_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_load_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>): (@pred_th_extract<mode>): Likewise. (*pred_th_extract<mode>): Likewise. * config/riscv/thead-vector-builtins-functions.def: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 +++ .../riscv/riscv-vector-builtins-shapes.cc | 160 +++++++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + gcc/config/riscv/riscv-vector-builtins.cc | 70 +++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 250 ++++++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 +++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 +++++ 15 files changed, 1104 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 1aa6e3c6665..b6f6e4ff37e 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2141,6 +2141,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2603,6 +2680,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> vsuxw_obj; +static CONSTEXPR const th_extract vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2894,6 +3002,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (vlb) +BASE (vlh) +BASE (vlw) +BASE (vlbu) +BASE (vlhu) +BASE (vlwu) +BASE (vsb) +BASE (vsh) +BASE (vsw) +BASE (vlsb) +BASE (vlsh) +BASE (vlsw) +BASE (vlsbu) +BASE (vlshu) +BASE (vlswu) +BASE (vssb) +BASE (vssh) +BASE (vssw) +BASE (vlxb) +BASE (vlxh) +BASE (vlxw) +BASE (vlxbu) +BASE (vlxhu) +BASE (vlxwu) +BASE (vsxb) +BASE (vsxh) +BASE (vsxw) +BASE (vsuxb) +BASE (vsuxh) +BASE (vsuxw) +BASE (vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 87c7f43a423..1f2c94d3541 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -280,6 +280,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const vlb; +extern const function_base *const vlh; +extern const function_base *const vlw; +extern const function_base *const vlbu; +extern const function_base *const vlhu; +extern const function_base *const vlwu; +extern const function_base *const vsb; +extern const function_base *const vsh; +extern const function_base *const vsw; +extern const function_base *const vlsb; +extern const function_base *const vlsh; +extern const function_base *const vlsw; +extern const function_base *const vlsbu; +extern const function_base *const vlshu; +extern const function_base *const vlswu; +extern const function_base *const vssb; +extern const function_base *const vssh; +extern const function_base *const vssw; +extern const function_base *const vlxb; +extern const function_base *const vlxh; +extern const function_base *const vlxw; +extern const function_base *const vlxbu; +extern const function_base *const vlxhu; +extern const function_base *const vlxwu; +extern const function_base *const vsxb; +extern const function_base *const vsxh; +extern const function_base *const vsxw; +extern const function_base *const vsuxb; +extern const function_base *const vsuxh; +extern const function_base *const vsuxw; +extern const function_base *const vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..8e90b17a94b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,146 @@ struct indexed_loadstore_def : public function_shape } }; +/* Add one function instance for GROUP, using operand suffix at index OI, + mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ +static void +build_th_loadstore (function_builder &b, const function_group_info &group, + unsigned int pred_idx, unsigned int vec_type_idx) +{ + auto_vec<tree, 5> argument_types; + function_instance function_instance (group.base_name, *group.base, + *group.shape, + group.ops_infos.types[vec_type_idx], + group.preds[pred_idx], &group.ops_infos); + tree return_type = group.ops_infos.ret.get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + b.allocate_argument_types (function_instance, argument_types); + b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + + tree type = builtin_types[group.ops_infos.types[vec_type_idx].index].vector; + if (strstr (group.base_name, "l") + && strstr (group.base_name, "u") + && !TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + if (strstr (group.base_name, "l") + && !strstr (group.base_name, "u") + && TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + machine_mode mode = TYPE_MODE (type); + int sew = GET_MODE_BITSIZE (GET_MODE_INNER (mode)); + if (strstr (group.base_name, "h") && sew == 8) + return; + + if (strstr (group.base_name, "w") && (sew == 8 || sew ==16)) + return; + + b.add_overloaded_function (function_instance, *group.shape); + b.add_unique_function (function_instance, (*group.shape), return_type, + argument_types); +} + +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +772,23 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1251,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1265,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..44b9fec1898 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -934,6 +934,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1481,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2672,38 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * index_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +2882,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..22fed60b4c3 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -123,6 +123,7 @@ enum required_ext ZVKNHB_EXT, /* Crypto vector Zvknhb sub-ext */ ZVKSED_EXT, /* Crypto vector Zvksed sub-ext */ ZVKSH_EXT, /* Crypto vector Zvksh sub-ext */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ }; /* Enumerates the RVV operand types. */ @@ -252,6 +253,8 @@ struct function_group_info return TARGET_ZVKSED; case ZVKSH_EXT: return TARGET_ZVKSH; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; default: gcc_unreachable (); } diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..fd3ba29bae9 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlhu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlwu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vsb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vlsb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlshu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlswu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vssb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vlxb, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxh, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxw, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxbu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxhu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxwu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vsxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..5fe9ba08c4e 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,165 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& riscv_vector::whole_reg_to_reg_move_p (operands, <MODE>mode, 7)" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..3c12c124597 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..30bef369375 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..3c8b5ccc16b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..b7c00404f18 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..17a53012acf --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..b187cfc852b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-11 10:54 ` Re：[PATCH " joshua @ 2024-01-11 10:55 ` juzhe.zhong 2024-01-11 14:11 ` Re：Re：[PATCH " joshua 0 siblings, 1 reply; 133+ messages in thread From: juzhe.zhong @ 2024-01-11 10:55 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 55958 bytes --] Yes. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 18:54 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Do you mean removing TARGET_XTHEADVECTOR for sext/zext patterns and then resending the "RISC-V: Handle differences between XTheadvector and Vector" patch? ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:57 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. LGTM. Could you resend the patch "RISC-V: Handle differences between XTheadvector and Vector Thanks. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-11 17:52 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (class th_extract): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (struct th_extract_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): * config/riscv/riscv-vector-builtins.h (enum required_ext): (struct function_group_info): * config/riscv/t-riscv: Add thead-vector-builtins-functions.def * config/riscv/thead-vector.md (@pred_mov_width<vlmem_op_attr><mode>): Add new patterns. (*pred_mov_width<vlmem_op_attr><mode>): Likewise. (@pred_store_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_load_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_store_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_load_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>): (@pred_th_extract<mode>): Likewise. (*pred_th_extract<mode>): Likewise. * config/riscv/thead-vector-builtins-functions.def: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 +++ .../riscv/riscv-vector-builtins-shapes.cc | 160 +++++++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + gcc/config/riscv/riscv-vector-builtins.cc | 70 +++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 250 ++++++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 +++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 +++++ 15 files changed, 1104 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 1aa6e3c6665..b6f6e4ff37e 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2141,6 +2141,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2603,6 +2680,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> vsuxw_obj; +static CONSTEXPR const th_extract vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2894,6 +3002,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (vlb) +BASE (vlh) +BASE (vlw) +BASE (vlbu) +BASE (vlhu) +BASE (vlwu) +BASE (vsb) +BASE (vsh) +BASE (vsw) +BASE (vlsb) +BASE (vlsh) +BASE (vlsw) +BASE (vlsbu) +BASE (vlshu) +BASE (vlswu) +BASE (vssb) +BASE (vssh) +BASE (vssw) +BASE (vlxb) +BASE (vlxh) +BASE (vlxw) +BASE (vlxbu) +BASE (vlxhu) +BASE (vlxwu) +BASE (vsxb) +BASE (vsxh) +BASE (vsxw) +BASE (vsuxb) +BASE (vsuxh) +BASE (vsuxw) +BASE (vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 87c7f43a423..1f2c94d3541 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -280,6 +280,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const vlb; +extern const function_base *const vlh; +extern const function_base *const vlw; +extern const function_base *const vlbu; +extern const function_base *const vlhu; +extern const function_base *const vlwu; +extern const function_base *const vsb; +extern const function_base *const vsh; +extern const function_base *const vsw; +extern const function_base *const vlsb; +extern const function_base *const vlsh; +extern const function_base *const vlsw; +extern const function_base *const vlsbu; +extern const function_base *const vlshu; +extern const function_base *const vlswu; +extern const function_base *const vssb; +extern const function_base *const vssh; +extern const function_base *const vssw; +extern const function_base *const vlxb; +extern const function_base *const vlxh; +extern const function_base *const vlxw; +extern const function_base *const vlxbu; +extern const function_base *const vlxhu; +extern const function_base *const vlxwu; +extern const function_base *const vsxb; +extern const function_base *const vsxh; +extern const function_base *const vsxw; +extern const function_base *const vsuxb; +extern const function_base *const vsuxh; +extern const function_base *const vsuxw; +extern const function_base *const vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..8e90b17a94b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,146 @@ struct indexed_loadstore_def : public function_shape } }; +/* Add one function instance for GROUP, using operand suffix at index OI, + mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ +static void +build_th_loadstore (function_builder &b, const function_group_info &group, + unsigned int pred_idx, unsigned int vec_type_idx) +{ + auto_vec<tree, 5> argument_types; + function_instance function_instance (group.base_name, *group.base, + *group.shape, + group.ops_infos.types[vec_type_idx], + group.preds[pred_idx], &group.ops_infos); + tree return_type = group.ops_infos.ret.get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + b.allocate_argument_types (function_instance, argument_types); + b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + + tree type = builtin_types[group.ops_infos.types[vec_type_idx].index].vector; + if (strstr (group.base_name, "l") + && strstr (group.base_name, "u") + && !TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + if (strstr (group.base_name, "l") + && !strstr (group.base_name, "u") + && TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + machine_mode mode = TYPE_MODE (type); + int sew = GET_MODE_BITSIZE (GET_MODE_INNER (mode)); + if (strstr (group.base_name, "h") && sew == 8) + return; + + if (strstr (group.base_name, "w") && (sew == 8 || sew ==16)) + return; + + b.add_overloaded_function (function_instance, *group.shape); + b.add_unique_function (function_instance, (*group.shape), return_type, + argument_types); +} + +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +772,23 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1251,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1265,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..44b9fec1898 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -934,6 +934,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1481,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2672,38 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * index_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +2882,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..22fed60b4c3 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -123,6 +123,7 @@ enum required_ext ZVKNHB_EXT, /* Crypto vector Zvknhb sub-ext */ ZVKSED_EXT, /* Crypto vector Zvksed sub-ext */ ZVKSH_EXT, /* Crypto vector Zvksh sub-ext */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ }; /* Enumerates the RVV operand types. */ @@ -252,6 +253,8 @@ struct function_group_info return TARGET_ZVKSED; case ZVKSH_EXT: return TARGET_ZVKSH; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; default: gcc_unreachable (); } diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..fd3ba29bae9 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlhu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlwu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vsb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vlsb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlshu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlswu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vssb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vlxb, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxh, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxw, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxbu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxhu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxwu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vsxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..5fe9ba08c4e 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,165 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& riscv_vector::whole_reg_to_reg_move_p (operands, <MODE>mode, 7)" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..3c12c124597 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..30bef369375 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..3c8b5ccc16b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..b7c00404f18 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..17a53012acf --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..b187cfc852b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-11 10:55 ` juzhe.zhong @ 2024-01-11 14:11 ` joshua 2024-01-11 22:59 ` Re：[PATCH " 钟居哲 2024-01-11 23:22 ` 钟居哲 0 siblings, 2 replies; 133+ messages in thread From: joshua @ 2024-01-11 14:11 UTC (permalink / raw) To: juzhe.zhong, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu Is the patch with !TARGET_XTHEADVECTOR for sext/zext patterns removed OK to commit? https://gcc.gnu.org/pipermail/gcc-patches/2024-January/642657.html ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 18:56 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Yes. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 18:54 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Do you mean removing TARGET_XTHEADVECTOR for sext/zext patterns and then resending the "RISC-V: Handle differences between XTheadvector and Vector" patch? ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:57 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. LGTM. Could you resend the patch "RISC-V: Handle differences between XTheadvector and Vector Thanks. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-11 17:52 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (class th_extract): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (struct th_extract_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): * config/riscv/riscv-vector-builtins.h (enum required_ext): (struct function_group_info): * config/riscv/t-riscv: Add thead-vector-builtins-functions.def * config/riscv/thead-vector.md (@pred_mov_width<vlmem_op_attr><mode>): Add new patterns. (*pred_mov_width<vlmem_op_attr><mode>): Likewise. (@pred_store_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_load_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_store_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_load_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>): (@pred_th_extract<mode>): Likewise. (*pred_th_extract<mode>): Likewise. * config/riscv/thead-vector-builtins-functions.def: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 +++ .../riscv/riscv-vector-builtins-shapes.cc | 160 +++++++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + gcc/config/riscv/riscv-vector-builtins.cc | 70 +++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 250 ++++++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 +++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 +++++ 15 files changed, 1104 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 1aa6e3c6665..b6f6e4ff37e 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2141,6 +2141,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2603,6 +2680,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> vsuxw_obj; +static CONSTEXPR const th_extract vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2894,6 +3002,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (vlb) +BASE (vlh) +BASE (vlw) +BASE (vlbu) +BASE (vlhu) +BASE (vlwu) +BASE (vsb) +BASE (vsh) +BASE (vsw) +BASE (vlsb) +BASE (vlsh) +BASE (vlsw) +BASE (vlsbu) +BASE (vlshu) +BASE (vlswu) +BASE (vssb) +BASE (vssh) +BASE (vssw) +BASE (vlxb) +BASE (vlxh) +BASE (vlxw) +BASE (vlxbu) +BASE (vlxhu) +BASE (vlxwu) +BASE (vsxb) +BASE (vsxh) +BASE (vsxw) +BASE (vsuxb) +BASE (vsuxh) +BASE (vsuxw) +BASE (vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 87c7f43a423..1f2c94d3541 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -280,6 +280,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const vlb; +extern const function_base *const vlh; +extern const function_base *const vlw; +extern const function_base *const vlbu; +extern const function_base *const vlhu; +extern const function_base *const vlwu; +extern const function_base *const vsb; +extern const function_base *const vsh; +extern const function_base *const vsw; +extern const function_base *const vlsb; +extern const function_base *const vlsh; +extern const function_base *const vlsw; +extern const function_base *const vlsbu; +extern const function_base *const vlshu; +extern const function_base *const vlswu; +extern const function_base *const vssb; +extern const function_base *const vssh; +extern const function_base *const vssw; +extern const function_base *const vlxb; +extern const function_base *const vlxh; +extern const function_base *const vlxw; +extern const function_base *const vlxbu; +extern const function_base *const vlxhu; +extern const function_base *const vlxwu; +extern const function_base *const vsxb; +extern const function_base *const vsxh; +extern const function_base *const vsxw; +extern const function_base *const vsuxb; +extern const function_base *const vsuxh; +extern const function_base *const vsuxw; +extern const function_base *const vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..8e90b17a94b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,146 @@ struct indexed_loadstore_def : public function_shape } }; +/* Add one function instance for GROUP, using operand suffix at index OI, + mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ +static void +build_th_loadstore (function_builder &b, const function_group_info &group, + unsigned int pred_idx, unsigned int vec_type_idx) +{ + auto_vec<tree, 5> argument_types; + function_instance function_instance (group.base_name, *group.base, + *group.shape, + group.ops_infos.types[vec_type_idx], + group.preds[pred_idx], &group.ops_infos); + tree return_type = group.ops_infos.ret.get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + b.allocate_argument_types (function_instance, argument_types); + b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + + tree type = builtin_types[group.ops_infos.types[vec_type_idx].index].vector; + if (strstr (group.base_name, "l") + && strstr (group.base_name, "u") + && !TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + if (strstr (group.base_name, "l") + && !strstr (group.base_name, "u") + && TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + machine_mode mode = TYPE_MODE (type); + int sew = GET_MODE_BITSIZE (GET_MODE_INNER (mode)); + if (strstr (group.base_name, "h") && sew == 8) + return; + + if (strstr (group.base_name, "w") && (sew == 8 || sew ==16)) + return; + + b.add_overloaded_function (function_instance, *group.shape); + b.add_unique_function (function_instance, (*group.shape), return_type, + argument_types); +} + +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +772,23 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1251,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1265,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..44b9fec1898 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -934,6 +934,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1481,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2672,38 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * index_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +2882,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..22fed60b4c3 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -123,6 +123,7 @@ enum required_ext ZVKNHB_EXT, /* Crypto vector Zvknhb sub-ext */ ZVKSED_EXT, /* Crypto vector Zvksed sub-ext */ ZVKSH_EXT, /* Crypto vector Zvksh sub-ext */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ }; /* Enumerates the RVV operand types. */ @@ -252,6 +253,8 @@ struct function_group_info return TARGET_ZVKSED; case ZVKSH_EXT: return TARGET_ZVKSH; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; default: gcc_unreachable (); } diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..fd3ba29bae9 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlhu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlwu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vsb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vlsb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlshu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlswu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vssb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vlxb, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxh, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxw, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxbu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxhu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxwu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vsxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..5fe9ba08c4e 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,165 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& riscv_vector::whole_reg_to_reg_move_p (operands, <MODE>mode, 7)" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..3c12c124597 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..30bef369375 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..3c8b5ccc16b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..b7c00404f18 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..17a53012acf --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..b187cfc852b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-11 14:11 ` Re：Re：[PATCH " joshua @ 2024-01-11 22:59 ` 钟居哲 2024-01-11 23:22 ` 钟居哲 1 sibling, 0 replies; 133+ messages in thread From: 钟居哲 @ 2024-01-11 22:59 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, Jeff Law, Christoph Müllner, jinma, Cooper Qu [-- Attachment #1: Type: text/plain, Size: 74442 bytes --] diff --git a/gcc/config/riscv/riscv-vector-switch.def b/gcc/config/riscv/riscv-vector-switch.def index 1ad26c2a3b2..452283b7416 100644 --- a/gcc/config/riscv/riscv-vector-switch.def +++ b/gcc/config/riscv/riscv-vector-switch.def @@ -68,9 +68,9 @@ Encode the ratio of SEW/LMUL into the mask types. #endif /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8HI, true, LMUL_8, 2) ENTRY (RVVM4HI, true, LMUL_4, 4) ENTRY (RVVM2HI, true, LMUL_2, 8) ENTRY (RVVM1HI, true, LMUL_1, 16) -ENTRY (RVVMF2HI, true, LMUL_F2, 32) -ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HI, !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_16. */ ENTRY (RVVM8HF, TARGET_VECTOR_ELEN_FP_16, LMUL_8, 2) ENTRY (RVVM4HF, TARGET_VECTOR_ELEN_FP_16, LMUL_4, 4) ENTRY (RVVM2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_2, 8) ENTRY (RVVM1HF, TARGET_VECTOR_ELEN_FP_16, LMUL_1, 16) -ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_F2, 32) -ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8SI, true, LMUL_8, 4) ENTRY (RVVM4SI, true, LMUL_4, 8) ENTRY (RVVM2SI, true, LMUL_2, 16) ENTRY (RVVM1SI, true, LMUL_1, 32) -ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_32. */ ENTRY (RVVM8SF, TARGET_VECTOR_ELEN_FP_32, LMUL_8, 4) ENTRY (RVVM4SF, TARGET_VECTOR_ELEN_FP_32, LMUL_4, 8) ENTRY (RVVM2SF, TARGET_VECTOR_ELEN_FP_32, LMUL_2, 16) ENTRY (RVVM1SF, TARGET_VECTOR_ELEN_FP_32, LMUL_1, 32) -ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if !TARGET_VECTOR_ELEN_64. */ ENTRY (RVVM8DI, TARGET_VECTOR_ELEN_64, LMUL_8, 8) @@ -140,127 +140,127 @@ ENTRY (RVVM1DF, TARGET_VECTOR_ELEN_FP_64, LMUL_1, 64) #endif TUPLE_ENTRY (RVVM1x8QI, true, RVVM1QI, 8, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x8QI, true, RVVMF2QI, 8, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x8QI, true, RVVMF4QI, 8, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 8, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x8QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 8, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x8QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 8, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 8, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x7QI, true, RVVM1QI, 7, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x7QI, true, RVVMF2QI, 7, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x7QI, true, RVVMF4QI, 7, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 7, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x7QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 7, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x7QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 7, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 7, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x6QI, true, RVVM1QI, 6, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x6QI, true, RVVMF2QI, 6, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x6QI, true, RVVMF4QI, 6, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 6, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x6QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 6, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x6QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 6, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 6, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x5QI, true, RVVM1QI, 5, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x5QI, true, RVVMF2QI, 5, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x5QI, true, RVVMF4QI, 5, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 5, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x5QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 5, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x5QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 5, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 5, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x4QI, true, RVVM2QI, 4, LMUL_2, 4) TUPLE_ENTRY (RVVM1x4QI, true, RVVM1QI, 4, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x4QI, true, RVVMF2QI, 4, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x4QI, true, RVVMF4QI, 4, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 4, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x4QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 4, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x4QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 4, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 4, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x3QI, true, RVVM2QI, 3, LMUL_2, 4) TUPLE_ENTRY (RVVM1x3QI, true, RVVM1QI, 3, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x3QI, true, RVVMF2QI, 3, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x3QI, true, RVVMF4QI, 3, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 3, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x3QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 3, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x3QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 3, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 3, LMUL_F8, 64) TUPLE_ENTRY (RVVM4x2QI, true, RVVM4QI, 2, LMUL_4, 2) TUPLE_ENTRY (RVVM2x2QI, true, RVVM2QI, 2, LMUL_2, 4) TUPLE_ENTRY (RVVM1x2QI, true, RVVM1QI, 2, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x2QI, true, RVVMF2QI, 2, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x2QI, true, RVVMF4QI, 2, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 2, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x2QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 2, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x2QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 2, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 2, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x8HI, true, RVVM1HI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HI, true, RVVMF2HI, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HI, true, RVVM1HI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HI, true, RVVMF2HI, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HI, true, RVVM1HI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HI, true, RVVMF2HI, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HI, true, RVVM1HI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HI, true, RVVMF2HI, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HI, true, RVVM2HI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HI, true, RVVM1HI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HI, true, RVVMF2HI, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HI, true, RVVM2HI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HI, true, RVVM1HI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HI, true, RVVMF2HI, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HI, true, RVVM4HI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HI, true, RVVM2HI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HI, true, RVVM1HI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HI, true, RVVMF2HI, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM4HF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8SI, true, RVVM1SI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SI, true, RVVM1SI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SI, true, RVVM1SI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SI, true, RVVM1SI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SI, true, RVVM2SI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SI, true, RVVM1SI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SI, true, RVVM2SI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SI, true, RVVM1SI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SI, true, RVVM4SI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SI, true, RVVM2SI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SI, true, RVVM1SI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM4SF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 2, LMUL_F2, 32) Remove these changes. Instead, you'd better blocks those modes in riscv_v_ext_mode_p /* Return true if it is either RVV vector mode or RVV tuple mode. */ static bool riscv_v_ext_mode_p (machine_mode mode) { if (TARGET_XTHEADVECTOR) { /* Block all fractional modes. */ unsigned int nf = riscv_vector::get_nf (mode); poly_uint64 mode_size = GET_MODE_SIZE (mode); if (nf > 1) mode_size = exact_div (mode_size, nf); if (known_lt (mode_size, BYTES_PER_RISCV_VECTOR)) return false; } return riscv_v_ext_vector_mode_p (mode) || riscv_v_ext_tuple_mode_p (mode) || riscv_v_ext_vls_mode_p (mode); } juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 22:11 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Is the patch with !TARGET_XTHEADVECTOR for sext/zext patterns removed OK to commit? https://gcc.gnu.org/pipermail/gcc-patches/2024-January/642657.html ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 18:56 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Yes. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 18:54 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Do you mean removing TARGET_XTHEADVECTOR for sext/zext patterns and then resending the "RISC-V: Handle differences between XTheadvector and Vector" patch? ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:57 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. LGTM. Could you resend the patch "RISC-V: Handle differences between XTheadvector and Vector Thanks. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-11 17:52 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (class th_extract): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (struct th_extract_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): * config/riscv/riscv-vector-builtins.h (enum required_ext): (struct function_group_info): * config/riscv/t-riscv: Add thead-vector-builtins-functions.def * config/riscv/thead-vector.md (@pred_mov_width<vlmem_op_attr><mode>): Add new patterns. (*pred_mov_width<vlmem_op_attr><mode>): Likewise. (@pred_store_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_load_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_store_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_load_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>): (@pred_th_extract<mode>): Likewise. (*pred_th_extract<mode>): Likewise. * config/riscv/thead-vector-builtins-functions.def: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 +++ .../riscv/riscv-vector-builtins-shapes.cc | 160 +++++++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + gcc/config/riscv/riscv-vector-builtins.cc | 70 +++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 250 ++++++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 +++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 +++++ 15 files changed, 1104 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 1aa6e3c6665..b6f6e4ff37e 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2141,6 +2141,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2603,6 +2680,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> vsuxw_obj; +static CONSTEXPR const th_extract vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2894,6 +3002,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (vlb) +BASE (vlh) +BASE (vlw) +BASE (vlbu) +BASE (vlhu) +BASE (vlwu) +BASE (vsb) +BASE (vsh) +BASE (vsw) +BASE (vlsb) +BASE (vlsh) +BASE (vlsw) +BASE (vlsbu) +BASE (vlshu) +BASE (vlswu) +BASE (vssb) +BASE (vssh) +BASE (vssw) +BASE (vlxb) +BASE (vlxh) +BASE (vlxw) +BASE (vlxbu) +BASE (vlxhu) +BASE (vlxwu) +BASE (vsxb) +BASE (vsxh) +BASE (vsxw) +BASE (vsuxb) +BASE (vsuxh) +BASE (vsuxw) +BASE (vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 87c7f43a423..1f2c94d3541 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -280,6 +280,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const vlb; +extern const function_base *const vlh; +extern const function_base *const vlw; +extern const function_base *const vlbu; +extern const function_base *const vlhu; +extern const function_base *const vlwu; +extern const function_base *const vsb; +extern const function_base *const vsh; +extern const function_base *const vsw; +extern const function_base *const vlsb; +extern const function_base *const vlsh; +extern const function_base *const vlsw; +extern const function_base *const vlsbu; +extern const function_base *const vlshu; +extern const function_base *const vlswu; +extern const function_base *const vssb; +extern const function_base *const vssh; +extern const function_base *const vssw; +extern const function_base *const vlxb; +extern const function_base *const vlxh; +extern const function_base *const vlxw; +extern const function_base *const vlxbu; +extern const function_base *const vlxhu; +extern const function_base *const vlxwu; +extern const function_base *const vsxb; +extern const function_base *const vsxh; +extern const function_base *const vsxw; +extern const function_base *const vsuxb; +extern const function_base *const vsuxh; +extern const function_base *const vsuxw; +extern const function_base *const vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..8e90b17a94b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,146 @@ struct indexed_loadstore_def : public function_shape } }; +/* Add one function instance for GROUP, using operand suffix at index OI, + mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ +static void +build_th_loadstore (function_builder &b, const function_group_info &group, + unsigned int pred_idx, unsigned int vec_type_idx) +{ + auto_vec<tree, 5> argument_types; + function_instance function_instance (group.base_name, *group.base, + *group.shape, + group.ops_infos.types[vec_type_idx], + group.preds[pred_idx], &group.ops_infos); + tree return_type = group.ops_infos.ret.get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + b.allocate_argument_types (function_instance, argument_types); + b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + + tree type = builtin_types[group.ops_infos.types[vec_type_idx].index].vector; + if (strstr (group.base_name, "l") + && strstr (group.base_name, "u") + && !TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + if (strstr (group.base_name, "l") + && !strstr (group.base_name, "u") + && TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + machine_mode mode = TYPE_MODE (type); + int sew = GET_MODE_BITSIZE (GET_MODE_INNER (mode)); + if (strstr (group.base_name, "h") && sew == 8) + return; + + if (strstr (group.base_name, "w") && (sew == 8 || sew ==16)) + return; + + b.add_overloaded_function (function_instance, *group.shape); + b.add_unique_function (function_instance, (*group.shape), return_type, + argument_types); +} + +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +772,23 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1251,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1265,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..44b9fec1898 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -934,6 +934,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1481,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2672,38 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * index_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +2882,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..22fed60b4c3 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -123,6 +123,7 @@ enum required_ext ZVKNHB_EXT, /* Crypto vector Zvknhb sub-ext */ ZVKSED_EXT, /* Crypto vector Zvksed sub-ext */ ZVKSH_EXT, /* Crypto vector Zvksh sub-ext */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ }; /* Enumerates the RVV operand types. */ @@ -252,6 +253,8 @@ struct function_group_info return TARGET_ZVKSED; case ZVKSH_EXT: return TARGET_ZVKSH; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; default: gcc_unreachable (); } diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..fd3ba29bae9 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlhu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlwu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vsb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vlsb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlshu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlswu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vssb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vlxb, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxh, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxw, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxbu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxhu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxwu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vsxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..5fe9ba08c4e 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,165 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& riscv_vector::whole_reg_to_reg_move_p (operands, <MODE>mode, 7)" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..3c12c124597 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..30bef369375 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..3c8b5ccc16b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..b7c00404f18 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..17a53012acf --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..b187cfc852b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-11 14:11 ` Re：Re：[PATCH " joshua 2024-01-11 22:59 ` Re：[PATCH " 钟居哲 @ 2024-01-11 23:22 ` 钟居哲 2024-01-12 0:49 ` 回复：Re：[PATCH " joshua 1 sibling, 1 reply; 133+ messages in thread From: 钟居哲 @ 2024-01-11 23:22 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, Jeff Law, Christoph Müllner, jinma, Cooper Qu [-- Attachment #1: Type: text/plain, Size: 60266 bytes --] Also, you don't need to change these: -ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) -ENTRY (RVVMF32BI, true, LMUL_F4, 32) -ENTRY (RVVMF16BI, true, LMUL_F2, 16) +ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F8, 64) +ENTRY (RVVMF32BI, true, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F4, 32) +ENTRY (RVVMF16BI, true, TARGET_XTHEADVECTOR ? LMUL_1 : LMUL_F2 , 16) You should specify these cases in enum vlmul_type get_vlmul (machine_mode mode) And don't change: (define_attr "sew" "" - (cond [(eq_attr "mode" "RVVMF64BI,RVVMF32BI,RVVMF16BI,RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ + (cond [(eq_attr "mode" "RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ RVVM8QI,RVVM4QI,RVVM2QI,RVVM1QI,RVVMF2QI,RVVMF4QI,RVVMF8QI,\ RVVM1x8QI,RVVMF2x8QI,RVVMF4x8QI,RVVMF8x8QI,\ RVVM1x7QI,RVVMF2x7QI,RVVMF4x7QI,RVVMF8x7QI,\ @@ -99,6 +99,18 @@ V1QI,V2QI,V4QI,V8QI,V16QI,V32QI,V64QI,V128QI,V256QI,V512QI,V1024QI,V2048QI,V4096QI,\ V1BI,V2BI,V4BI,V8BI,V16BI,V32BI,V64BI,V128BI,V256BI,V512BI,V1024BI,V2048BI,V4096BI") (const_int 8) + (eq_attr "mode" "RVVMF16BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 16) + (const_int 8)) + (eq_attr "mode" "RVVMF32BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 32) + (const_int 8)) + (eq_attr "mode" "RVVMF64BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 64) + (const_int 8)) (eq_attr "mode" "RVVM8HI,RVVM4HI,RVVM2HI,RVVM1HI,RVVMF2HI,RVVMF4HI,\ RVVM1x8HI,RVVMF2x8HI,RVVMF4x8HI,\ RVVM1x7HI,RVVMF2x7HI,RVVMF4x7HI,\ @@ -159,9 +171,9 @@ (eq_attr "mode" "RVVM4QI,RVVMF2BI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2QI,RVVMF4BI") (symbol_ref "riscv_vector::LMUL_2") (eq_attr "mode" "RVVM1QI,RVVMF8BI") (symbol_ref "riscv_vector::LMUL_1") - (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "riscv_vector::LMUL_F2") - (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "riscv_vector::LMUL_F4") - (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "riscv_vector::LMUL_F8") + (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F2") + (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F4") + (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F8")You should change /* Helper function to get SEW operand. We always have SEW value for all RVV instructions that have VTYPE OP. */ uint8_t get_sew (rtx_insn *rinsn) { return get_attr_sew (rinsn); } /* Helper function to get VLMUL operand. We always have VLMUL value for all RVV instructions that have VTYPE OP. */ enum vlmul_type get_vlmul (rtx_insn *rinsn) { return (enum vlmul_type) get_attr_vlmul (rinsn); } + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE)Why do you add this ? juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 22:11 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Is the patch with !TARGET_XTHEADVECTOR for sext/zext patterns removed OK to commit? https://gcc.gnu.org/pipermail/gcc-patches/2024-January/642657.html ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 18:56 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Yes. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 18:54 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Do you mean removing TARGET_XTHEADVECTOR for sext/zext patterns and then resending the "RISC-V: Handle differences between XTheadvector and Vector" patch? ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:57 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. LGTM. Could you resend the patch "RISC-V: Handle differences between XTheadvector and Vector Thanks. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-11 17:52 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (class th_extract): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (struct th_extract_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): * config/riscv/riscv-vector-builtins.h (enum required_ext): (struct function_group_info): * config/riscv/t-riscv: Add thead-vector-builtins-functions.def * config/riscv/thead-vector.md (@pred_mov_width<vlmem_op_attr><mode>): Add new patterns. (*pred_mov_width<vlmem_op_attr><mode>): Likewise. (@pred_store_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_load_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_store_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_load_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>): (@pred_th_extract<mode>): Likewise. (*pred_th_extract<mode>): Likewise. * config/riscv/thead-vector-builtins-functions.def: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 +++ .../riscv/riscv-vector-builtins-shapes.cc | 160 +++++++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + gcc/config/riscv/riscv-vector-builtins.cc | 70 +++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 250 ++++++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 +++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 +++++ 15 files changed, 1104 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 1aa6e3c6665..b6f6e4ff37e 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2141,6 +2141,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2603,6 +2680,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> vsuxw_obj; +static CONSTEXPR const th_extract vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2894,6 +3002,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (vlb) +BASE (vlh) +BASE (vlw) +BASE (vlbu) +BASE (vlhu) +BASE (vlwu) +BASE (vsb) +BASE (vsh) +BASE (vsw) +BASE (vlsb) +BASE (vlsh) +BASE (vlsw) +BASE (vlsbu) +BASE (vlshu) +BASE (vlswu) +BASE (vssb) +BASE (vssh) +BASE (vssw) +BASE (vlxb) +BASE (vlxh) +BASE (vlxw) +BASE (vlxbu) +BASE (vlxhu) +BASE (vlxwu) +BASE (vsxb) +BASE (vsxh) +BASE (vsxw) +BASE (vsuxb) +BASE (vsuxh) +BASE (vsuxw) +BASE (vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 87c7f43a423..1f2c94d3541 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -280,6 +280,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const vlb; +extern const function_base *const vlh; +extern const function_base *const vlw; +extern const function_base *const vlbu; +extern const function_base *const vlhu; +extern const function_base *const vlwu; +extern const function_base *const vsb; +extern const function_base *const vsh; +extern const function_base *const vsw; +extern const function_base *const vlsb; +extern const function_base *const vlsh; +extern const function_base *const vlsw; +extern const function_base *const vlsbu; +extern const function_base *const vlshu; +extern const function_base *const vlswu; +extern const function_base *const vssb; +extern const function_base *const vssh; +extern const function_base *const vssw; +extern const function_base *const vlxb; +extern const function_base *const vlxh; +extern const function_base *const vlxw; +extern const function_base *const vlxbu; +extern const function_base *const vlxhu; +extern const function_base *const vlxwu; +extern const function_base *const vsxb; +extern const function_base *const vsxh; +extern const function_base *const vsxw; +extern const function_base *const vsuxb; +extern const function_base *const vsuxh; +extern const function_base *const vsuxw; +extern const function_base *const vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..8e90b17a94b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,146 @@ struct indexed_loadstore_def : public function_shape } }; +/* Add one function instance for GROUP, using operand suffix at index OI, + mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ +static void +build_th_loadstore (function_builder &b, const function_group_info &group, + unsigned int pred_idx, unsigned int vec_type_idx) +{ + auto_vec<tree, 5> argument_types; + function_instance function_instance (group.base_name, *group.base, + *group.shape, + group.ops_infos.types[vec_type_idx], + group.preds[pred_idx], &group.ops_infos); + tree return_type = group.ops_infos.ret.get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + b.allocate_argument_types (function_instance, argument_types); + b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + + tree type = builtin_types[group.ops_infos.types[vec_type_idx].index].vector; + if (strstr (group.base_name, "l") + && strstr (group.base_name, "u") + && !TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + if (strstr (group.base_name, "l") + && !strstr (group.base_name, "u") + && TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + machine_mode mode = TYPE_MODE (type); + int sew = GET_MODE_BITSIZE (GET_MODE_INNER (mode)); + if (strstr (group.base_name, "h") && sew == 8) + return; + + if (strstr (group.base_name, "w") && (sew == 8 || sew ==16)) + return; + + b.add_overloaded_function (function_instance, *group.shape); + b.add_unique_function (function_instance, (*group.shape), return_type, + argument_types); +} + +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +772,23 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1251,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1265,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..44b9fec1898 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -934,6 +934,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1481,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2672,38 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * index_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +2882,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..22fed60b4c3 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -123,6 +123,7 @@ enum required_ext ZVKNHB_EXT, /* Crypto vector Zvknhb sub-ext */ ZVKSED_EXT, /* Crypto vector Zvksed sub-ext */ ZVKSH_EXT, /* Crypto vector Zvksh sub-ext */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ }; /* Enumerates the RVV operand types. */ @@ -252,6 +253,8 @@ struct function_group_info return TARGET_ZVKSED; case ZVKSH_EXT: return TARGET_ZVKSH; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; default: gcc_unreachable (); } diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..fd3ba29bae9 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlhu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlwu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vsb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vlsb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlshu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlswu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vssb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vlxb, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxh, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxw, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxbu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxhu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxwu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vsxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..5fe9ba08c4e 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,165 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& riscv_vector::whole_reg_to_reg_move_p (operands, <MODE>mode, 7)" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..3c12c124597 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..30bef369375 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..3c8b5ccc16b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..b7c00404f18 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..17a53012acf --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..b187cfc852b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* 回复：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-11 23:22 ` 钟居哲 @ 2024-01-12 0:49 ` joshua 2024-01-12 1:08 ` 回复：[PATCH " juzhe.zhong 2024-01-12 1:14 ` juzhe.zhong 0 siblings, 2 replies; 133+ messages in thread From: joshua @ 2024-01-12 0:49 UTC (permalink / raw) To: 钟居哲, gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, Jeff Law, Christoph Müllner, jinma, Cooper Qu [-- Attachment #1: Type: text/plain, Size: 58514 bytes --] The changes to lmul and sew are only done for vbool, which is RVVMF64BI/RVVMF32BI/RVVMF16BI, in the intrinsics with mask. That is because theadvector has no load/store mask instructions. lmul and sew in other non-bool types do not need to be modified. I don’t think directly change get_sew and get_vlmul is reasonable. + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE)Why do you add this ? theadvector does not have sew in load/storeinstructions. In O2, vector will remove onevsetvl for two consecutive vle with differentsew. However, things will go wrong under thisoptimization for theadvector. ------------------------------------------------------------------ 发件人：钟居哲 <juzhe.zhong@rivai.ai> 发送时间：2024年1月12日(星期五) 07:22 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送："jim.wilson.gcc"<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; Jeff Law<jeffreyalaw@gmail.com>; "Christoph Müllner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; Cooper Qu<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Also, you don't need to change these: -ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) -ENTRY (RVVMF32BI, true, LMUL_F4, 32) -ENTRY (RVVMF16BI, true, LMUL_F2, 16) +ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F8, 64) +ENTRY (RVVMF32BI, true, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F4, 32) +ENTRY (RVVMF16BI, true, TARGET_XTHEADVECTOR ? LMUL_1 : LMUL_F2 , 16) You should specify these cases in enum vlmul_type get_vlmul (machine_mode mode) And don't change: (define_attr "sew" "" - (cond [(eq_attr "mode" "RVVMF64BI,RVVMF32BI,RVVMF16BI,RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ + (cond [(eq_attr "mode" "RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ RVVM8QI,RVVM4QI,RVVM2QI,RVVM1QI,RVVMF2QI,RVVMF4QI,RVVMF8QI,\ RVVM1x8QI,RVVMF2x8QI,RVVMF4x8QI,RVVMF8x8QI,\ RVVM1x7QI,RVVMF2x7QI,RVVMF4x7QI,RVVMF8x7QI,\ @@ -99,6 +99,18 @@ V1QI,V2QI,V4QI,V8QI,V16QI,V32QI,V64QI,V128QI,V256QI,V512QI,V1024QI,V2048QI,V4096QI,\ V1BI,V2BI,V4BI,V8BI,V16BI,V32BI,V64BI,V128BI,V256BI,V512BI,V1024BI,V2048BI,V4096BI") (const_int 8) + (eq_attr "mode" "RVVMF16BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 16) + (const_int 8)) + (eq_attr "mode" "RVVMF32BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 32) + (const_int 8)) + (eq_attr "mode" "RVVMF64BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 64) + (const_int 8)) (eq_attr "mode" "RVVM8HI,RVVM4HI,RVVM2HI,RVVM1HI,RVVMF2HI,RVVMF4HI,\ RVVM1x8HI,RVVMF2x8HI,RVVMF4x8HI,\ RVVM1x7HI,RVVMF2x7HI,RVVMF4x7HI,\ @@ -159,9 +171,9 @@ (eq_attr "mode" "RVVM4QI,RVVMF2BI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2QI,RVVMF4BI") (symbol_ref "riscv_vector::LMUL_2") (eq_attr "mode" "RVVM1QI,RVVMF8BI") (symbol_ref "riscv_vector::LMUL_1") - (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "riscv_vector::LMUL_F2") - (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "riscv_vector::LMUL_F4") - (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "riscv_vector::LMUL_F8") + (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F2") + (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F4") + (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F8")You should change /* Helper function to get SEW operand. We always have SEW value for all RVV instructions that have VTYPE OP. */ uint8_t get_sew (rtx_insn *rinsn) { return get_attr_sew (rinsn); } /* Helper function to get VLMUL operand. We always have VLMUL value for all RVV instructions that have VTYPE OP. */ enum vlmul_type get_vlmul (rtx_insn *rinsn) { return (enum vlmul_type) get_attr_vlmul (rinsn); } + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE)Why do you add this ? juzhe.zhong@rivai.ai 发件人： joshua <mailto:cooper.joshua@linux.alibaba.com > 发送时间： 2024-01-11 22:11 收件人： juzhe.zhong@rivai.ai <mailto:juzhe.zhong@rivai.ai >; gcc-patches <mailto:gcc-patches@gcc.gnu.org > 抄送： Jim Wilson <mailto:jim.wilson.gcc@gmail.com >; palmer <mailto:palmer@dabbelt.com >; andrew <mailto:andrew@sifive.com >; philipp.tomsich <mailto:philipp.tomsich@vrull.eu >; jeffreyalaw <mailto:jeffreyalaw@gmail.com >; christoph.muellner <mailto:christoph.muellner@vrull.eu >; jinma <mailto:jinma@linux.alibaba.com >; cooper.qu <mailto:cooper.qu@linux.alibaba.com > 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Is the patch with !TARGET_XTHEADVECTOR for sext/zext patterns removed OK to commit? https://gcc.gnu.org/pipermail/gcc-patches/2024-January/642657.html <https://gcc.gnu.org/pipermail/gcc-patches/2024-January/642657.html > ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 18:56 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Yes. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 18:54 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Do you mean removing TARGET_XTHEADVECTOR for sext/zext patterns and then resending the "RISC-V: Handle differences between XTheadvector and Vector" patch? ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:57 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. LGTM. Could you resend the patch "RISC-V: Handle differences between XTheadvector and Vector Thanks. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-11 17:52 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (class th_extract): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (struct th_extract_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): * config/riscv/riscv-vector-builtins.h (enum required_ext): (struct function_group_info): * config/riscv/t-riscv: Add thead-vector-builtins-functions.def * config/riscv/thead-vector.md (@pred_mov_width<vlmem_op_attr><mode>): Add new patterns. (*pred_mov_width<vlmem_op_attr><mode>): Likewise. (@pred_store_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_load_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_store_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_load_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>): (@pred_th_extract<mode>): Likewise. (*pred_th_extract<mode>): Likewise. * config/riscv/thead-vector-builtins-functions.def: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 +++ .../riscv/riscv-vector-builtins-shapes.cc | 160 +++++++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + gcc/config/riscv/riscv-vector-builtins.cc | 70 +++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 250 ++++++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 +++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 +++++ 15 files changed, 1104 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 1aa6e3c6665..b6f6e4ff37e 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2141,6 +2141,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2603,6 +2680,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> vsuxw_obj; +static CONSTEXPR const th_extract vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2894,6 +3002,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (vlb) +BASE (vlh) +BASE (vlw) +BASE (vlbu) +BASE (vlhu) +BASE (vlwu) +BASE (vsb) +BASE (vsh) +BASE (vsw) +BASE (vlsb) +BASE (vlsh) +BASE (vlsw) +BASE (vlsbu) +BASE (vlshu) +BASE (vlswu) +BASE (vssb) +BASE (vssh) +BASE (vssw) +BASE (vlxb) +BASE (vlxh) +BASE (vlxw) +BASE (vlxbu) +BASE (vlxhu) +BASE (vlxwu) +BASE (vsxb) +BASE (vsxh) +BASE (vsxw) +BASE (vsuxb) +BASE (vsuxh) +BASE (vsuxw) +BASE (vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 87c7f43a423..1f2c94d3541 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -280,6 +280,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const vlb; +extern const function_base *const vlh; +extern const function_base *const vlw; +extern const function_base *const vlbu; +extern const function_base *const vlhu; +extern const function_base *const vlwu; +extern const function_base *const vsb; +extern const function_base *const vsh; +extern const function_base *const vsw; +extern const function_base *const vlsb; +extern const function_base *const vlsh; +extern const function_base *const vlsw; +extern const function_base *const vlsbu; +extern const function_base *const vlshu; +extern const function_base *const vlswu; +extern const function_base *const vssb; +extern const function_base *const vssh; +extern const function_base *const vssw; +extern const function_base *const vlxb; +extern const function_base *const vlxh; +extern const function_base *const vlxw; +extern const function_base *const vlxbu; +extern const function_base *const vlxhu; +extern const function_base *const vlxwu; +extern const function_base *const vsxb; +extern const function_base *const vsxh; +extern const function_base *const vsxw; +extern const function_base *const vsuxb; +extern const function_base *const vsuxh; +extern const function_base *const vsuxw; +extern const function_base *const vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..8e90b17a94b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,146 @@ struct indexed_loadstore_def : public function_shape } }; +/* Add one function instance for GROUP, using operand suffix at index OI, + mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ +static void +build_th_loadstore (function_builder &b, const function_group_info &group, + unsigned int pred_idx, unsigned int vec_type_idx) +{ + auto_vec<tree, 5> argument_types; + function_instance function_instance (group.base_name, *group.base, + *group.shape, + group.ops_infos.types[vec_type_idx], + group.preds[pred_idx], &group.ops_infos); + tree return_type = group.ops_infos.ret.get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + b.allocate_argument_types (function_instance, argument_types); + b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + + tree type = builtin_types[group.ops_infos.types[vec_type_idx].index].vector; + if (strstr (group.base_name, "l") + && strstr (group.base_name, "u") + && !TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + if (strstr (group.base_name, "l") + && !strstr (group.base_name, "u") + && TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + machine_mode mode = TYPE_MODE (type); + int sew = GET_MODE_BITSIZE (GET_MODE_INNER (mode)); + if (strstr (group.base_name, "h") && sew == 8) + return; + + if (strstr (group.base_name, "w") && (sew == 8 || sew ==16)) + return; + + b.add_overloaded_function (function_instance, *group.shape); + b.add_unique_function (function_instance, (*group.shape), return_type, + argument_types); +} + +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +772,23 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1251,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1265,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..44b9fec1898 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -934,6 +934,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1481,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2672,38 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * index_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +2882,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..22fed60b4c3 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -123,6 +123,7 @@ enum required_ext ZVKNHB_EXT, /* Crypto vector Zvknhb sub-ext */ ZVKSED_EXT, /* Crypto vector Zvksed sub-ext */ ZVKSH_EXT, /* Crypto vector Zvksh sub-ext */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ }; /* Enumerates the RVV operand types. */ @@ -252,6 +253,8 @@ struct function_group_info return TARGET_ZVKSED; case ZVKSH_EXT: return TARGET_ZVKSH; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; default: gcc_unreachable (); } diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..fd3ba29bae9 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlhu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlwu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vsb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vlsb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlshu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlswu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vssb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vlxb, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxh, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxw, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxbu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxhu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxwu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vsxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..5fe9ba08c4e 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,165 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& riscv_vector::whole_reg_to_reg_move_p (operands, <MODE>mode, 7)" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..3c12c124597 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..30bef369375 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..3c8b5ccc16b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..b7c00404f18 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..17a53012acf --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..b187cfc852b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: 回复：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-12 0:49 ` 回复：Re：[PATCH " joshua @ 2024-01-12 1:08 ` juzhe.zhong 2024-01-12 1:14 ` juzhe.zhong 1 sibling, 0 replies; 133+ messages in thread From: juzhe.zhong @ 2024-01-12 1:08 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 61969 bytes --] Ok. Make sense to me. LGTM as long as you removed !TARGET_THEADVECTOR in sext/zext patterns. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-12 08:49 收件人： 钟居哲; gcc-patches 抄送： jim.wilson.gcc; palmer; andrew; philipp.tomsich; Jeff Law; Christoph Müllner; jinma; Cooper Qu 主题： 回复：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. The changes to lmul and sew are only done for vbool, which is RVVMF64BI/RVVMF32BI/RVVMF16BI, in the intrinsics with mask. That is because theadvector has no load/store mask instructions. lmul and sew in other non-bool types do not need to be modified. I don’t think directly change get_sew and get_vlmul is reasonable. + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE)Why do you add this ? theadvector does not have sew in load/storeinstructions. In O2, vector will remove onevsetvl for two consecutive vle with differentsew. However, things will go wrong under thisoptimization for theadvector. ------------------------------------------------------------------ 发件人：钟居哲 <juzhe.zhong@rivai.ai> 发送时间：2024年1月12日(星期五) 07:22 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送："jim.wilson.gcc"<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; Jeff Law<jeffreyalaw@gmail.com>; "Christoph Müllner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; Cooper Qu<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Also, you don't need to change these: -ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) -ENTRY (RVVMF32BI, true, LMUL_F4, 32) -ENTRY (RVVMF16BI, true, LMUL_F2, 16) +ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F8, 64) +ENTRY (RVVMF32BI, true, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F4, 32) +ENTRY (RVVMF16BI, true, TARGET_XTHEADVECTOR ? LMUL_1 : LMUL_F2 , 16) You should specify these cases in enum vlmul_type get_vlmul (machine_mode mode) And don't change: (define_attr "sew" "" - (cond [(eq_attr "mode" "RVVMF64BI,RVVMF32BI,RVVMF16BI,RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ + (cond [(eq_attr "mode" "RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ RVVM8QI,RVVM4QI,RVVM2QI,RVVM1QI,RVVMF2QI,RVVMF4QI,RVVMF8QI,\ RVVM1x8QI,RVVMF2x8QI,RVVMF4x8QI,RVVMF8x8QI,\ RVVM1x7QI,RVVMF2x7QI,RVVMF4x7QI,RVVMF8x7QI,\ @@ -99,6 +99,18 @@ V1QI,V2QI,V4QI,V8QI,V16QI,V32QI,V64QI,V128QI,V256QI,V512QI,V1024QI,V2048QI,V4096QI,\ V1BI,V2BI,V4BI,V8BI,V16BI,V32BI,V64BI,V128BI,V256BI,V512BI,V1024BI,V2048BI,V4096BI") (const_int 8) + (eq_attr "mode" "RVVMF16BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 16) + (const_int 8)) + (eq_attr "mode" "RVVMF32BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 32) + (const_int 8)) + (eq_attr "mode" "RVVMF64BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 64) + (const_int 8)) (eq_attr "mode" "RVVM8HI,RVVM4HI,RVVM2HI,RVVM1HI,RVVMF2HI,RVVMF4HI,\ RVVM1x8HI,RVVMF2x8HI,RVVMF4x8HI,\ RVVM1x7HI,RVVMF2x7HI,RVVMF4x7HI,\ @@ -159,9 +171,9 @@ (eq_attr "mode" "RVVM4QI,RVVMF2BI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2QI,RVVMF4BI") (symbol_ref "riscv_vector::LMUL_2") (eq_attr "mode" "RVVM1QI,RVVMF8BI") (symbol_ref "riscv_vector::LMUL_1") - (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "riscv_vector::LMUL_F2") - (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "riscv_vector::LMUL_F4") - (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "riscv_vector::LMUL_F8") + (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F2") + (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F4") + (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F8")You should change /* Helper function to get SEW operand. We always have SEW value for all RVV instructions that have VTYPE OP. */ uint8_t get_sew (rtx_insn *rinsn) { return get_attr_sew (rinsn); } /* Helper function to get VLMUL operand. We always have VLMUL value for all RVV instructions that have VTYPE OP. */ enum vlmul_type get_vlmul (rtx_insn *rinsn) { return (enum vlmul_type) get_attr_vlmul (rinsn); } + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE)Why do you add this ? juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 22:11 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Is the patch with !TARGET_XTHEADVECTOR for sext/zext patterns removed OK to commit? https://gcc.gnu.org/pipermail/gcc-patches/2024-January/642657.html ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 18:56 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Yes. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 18:54 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Do you mean removing TARGET_XTHEADVECTOR for sext/zext patterns and then resending the "RISC-V: Handle differences between XTheadvector and Vector" patch? ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:57 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. LGTM. Could you resend the patch "RISC-V: Handle differences between XTheadvector and Vector Thanks. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-11 17:52 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (class th_extract): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (struct th_extract_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): * config/riscv/riscv-vector-builtins.h (enum required_ext): (struct function_group_info): * config/riscv/t-riscv: Add thead-vector-builtins-functions.def * config/riscv/thead-vector.md (@pred_mov_width<vlmem_op_attr><mode>): Add new patterns. (*pred_mov_width<vlmem_op_attr><mode>): Likewise. (@pred_store_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_load_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_store_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_load_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>): (@pred_th_extract<mode>): Likewise. (*pred_th_extract<mode>): Likewise. * config/riscv/thead-vector-builtins-functions.def: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 +++ .../riscv/riscv-vector-builtins-shapes.cc | 160 +++++++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + gcc/config/riscv/riscv-vector-builtins.cc | 70 +++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 250 ++++++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 +++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 +++++ 15 files changed, 1104 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 1aa6e3c6665..b6f6e4ff37e 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2141,6 +2141,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2603,6 +2680,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> vsuxw_obj; +static CONSTEXPR const th_extract vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2894,6 +3002,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (vlb) +BASE (vlh) +BASE (vlw) +BASE (vlbu) +BASE (vlhu) +BASE (vlwu) +BASE (vsb) +BASE (vsh) +BASE (vsw) +BASE (vlsb) +BASE (vlsh) +BASE (vlsw) +BASE (vlsbu) +BASE (vlshu) +BASE (vlswu) +BASE (vssb) +BASE (vssh) +BASE (vssw) +BASE (vlxb) +BASE (vlxh) +BASE (vlxw) +BASE (vlxbu) +BASE (vlxhu) +BASE (vlxwu) +BASE (vsxb) +BASE (vsxh) +BASE (vsxw) +BASE (vsuxb) +BASE (vsuxh) +BASE (vsuxw) +BASE (vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 87c7f43a423..1f2c94d3541 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -280,6 +280,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const vlb; +extern const function_base *const vlh; +extern const function_base *const vlw; +extern const function_base *const vlbu; +extern const function_base *const vlhu; +extern const function_base *const vlwu; +extern const function_base *const vsb; +extern const function_base *const vsh; +extern const function_base *const vsw; +extern const function_base *const vlsb; +extern const function_base *const vlsh; +extern const function_base *const vlsw; +extern const function_base *const vlsbu; +extern const function_base *const vlshu; +extern const function_base *const vlswu; +extern const function_base *const vssb; +extern const function_base *const vssh; +extern const function_base *const vssw; +extern const function_base *const vlxb; +extern const function_base *const vlxh; +extern const function_base *const vlxw; +extern const function_base *const vlxbu; +extern const function_base *const vlxhu; +extern const function_base *const vlxwu; +extern const function_base *const vsxb; +extern const function_base *const vsxh; +extern const function_base *const vsxw; +extern const function_base *const vsuxb; +extern const function_base *const vsuxh; +extern const function_base *const vsuxw; +extern const function_base *const vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..8e90b17a94b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,146 @@ struct indexed_loadstore_def : public function_shape } }; +/* Add one function instance for GROUP, using operand suffix at index OI, + mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ +static void +build_th_loadstore (function_builder &b, const function_group_info &group, + unsigned int pred_idx, unsigned int vec_type_idx) +{ + auto_vec<tree, 5> argument_types; + function_instance function_instance (group.base_name, *group.base, + *group.shape, + group.ops_infos.types[vec_type_idx], + group.preds[pred_idx], &group.ops_infos); + tree return_type = group.ops_infos.ret.get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + b.allocate_argument_types (function_instance, argument_types); + b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + + tree type = builtin_types[group.ops_infos.types[vec_type_idx].index].vector; + if (strstr (group.base_name, "l") + && strstr (group.base_name, "u") + && !TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + if (strstr (group.base_name, "l") + && !strstr (group.base_name, "u") + && TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + machine_mode mode = TYPE_MODE (type); + int sew = GET_MODE_BITSIZE (GET_MODE_INNER (mode)); + if (strstr (group.base_name, "h") && sew == 8) + return; + + if (strstr (group.base_name, "w") && (sew == 8 || sew ==16)) + return; + + b.add_overloaded_function (function_instance, *group.shape); + b.add_unique_function (function_instance, (*group.shape), return_type, + argument_types); +} + +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +772,23 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1251,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1265,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..44b9fec1898 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -934,6 +934,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1481,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2672,38 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * index_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +2882,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..22fed60b4c3 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -123,6 +123,7 @@ enum required_ext ZVKNHB_EXT, /* Crypto vector Zvknhb sub-ext */ ZVKSED_EXT, /* Crypto vector Zvksed sub-ext */ ZVKSH_EXT, /* Crypto vector Zvksh sub-ext */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ }; /* Enumerates the RVV operand types. */ @@ -252,6 +253,8 @@ struct function_group_info return TARGET_ZVKSED; case ZVKSH_EXT: return TARGET_ZVKSH; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; default: gcc_unreachable (); } diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..fd3ba29bae9 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlhu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlwu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vsb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vlsb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlshu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlswu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vssb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vlxb, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxh, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxw, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxbu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxhu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxwu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vsxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..5fe9ba08c4e 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,165 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& riscv_vector::whole_reg_to_reg_move_p (operands, <MODE>mode, 7)" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..3c12c124597 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..30bef369375 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..3c8b5ccc16b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..b7c00404f18 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..17a53012acf --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..b187cfc852b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: 回复：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-12 0:49 ` 回复：Re：[PATCH " joshua 2024-01-12 1:08 ` 回复：[PATCH " juzhe.zhong @ 2024-01-12 1:14 ` juzhe.zhong 2024-01-12 3:26 ` Re：Re：[PATCH " joshua 1 sibling, 1 reply; 133+ messages in thread From: juzhe.zhong @ 2024-01-12 1:14 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 63594 bytes --] I notice the changelog is incorrect. Plz update the change log: gcc/ChangeLog: * config.gcc: Add files for XTheadVector intrinsics. * config/riscv/autovec.md: Guard XTheadVector. * config/riscv/riscv-c.cc: Add pragma for XTheadVector. * config/riscv/riscv-string.cc (expand_block_move): Guard XTheadVector. * config/riscv/riscv-string.cc (vls_mode_valid_p): Avoid autovec. * config/riscv/riscv-vector-builtins-shapes.cc (check_type): (build_one): New function. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): (DEF_THEAD_RVV_FUNCTION): Add new marcos. (check_required_extensions): (handle_pragma_vector): * config/riscv/riscv-vector-builtins.h (RVV_REQUIRE_VECTOR): (RVV_REQUIRE_XTHEADVECTOR): Add RVV_REQUIRE_VECTOR and RVV_REQUIRE_XTHEADVECTOR. (struct function_group_info): * config/riscv/riscv-vector-switch.def (ENTRY): Disable fractional mode for the XTheadVector extension. (TUPLE_ENTRY): Likewise. * config/riscv/riscv-vsetvl.cc: Add functions for xtheadvector. * config/riscv/riscv.cc (riscv_v_ext_vls_mode_p): Guard XTheadVector. (riscv_v_adjust_bytesize): Likewise. (riscv_preferred_simd_mode): Likewsie. (riscv_autovectorize_vector_modes): Likewise. (riscv_vector_mode_supported_any_target_p): Likewise. (TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P): Likewise. * config/riscv/vector.md: Include thead-vector.md. * config/riscv/riscv_th_vector.h: New file. * config/riscv/thead-vector.md: New file. The changlog doesn't match your patch. I suspect you other patches have same issues. Plz update them and resend. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-12 08:49 收件人： 钟居哲; gcc-patches 抄送： jim.wilson.gcc; palmer; andrew; philipp.tomsich; Jeff Law; Christoph Müllner; jinma; Cooper Qu 主题： 回复：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. The changes to lmul and sew are only done for vbool, which is RVVMF64BI/RVVMF32BI/RVVMF16BI, in the intrinsics with mask. That is because theadvector has no load/store mask instructions. lmul and sew in other non-bool types do not need to be modified. I don’t think directly change get_sew and get_vlmul is reasonable. + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE)Why do you add this ? theadvector does not have sew in load/storeinstructions. In O2, vector will remove onevsetvl for two consecutive vle with differentsew. However, things will go wrong under thisoptimization for theadvector. ------------------------------------------------------------------ 发件人：钟居哲 <juzhe.zhong@rivai.ai> 发送时间：2024年1月12日(星期五) 07:22 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送："jim.wilson.gcc"<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; Jeff Law<jeffreyalaw@gmail.com>; "Christoph Müllner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; Cooper Qu<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Also, you don't need to change these: -ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) -ENTRY (RVVMF32BI, true, LMUL_F4, 32) -ENTRY (RVVMF16BI, true, LMUL_F2, 16) +ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F8, 64) +ENTRY (RVVMF32BI, true, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F4, 32) +ENTRY (RVVMF16BI, true, TARGET_XTHEADVECTOR ? LMUL_1 : LMUL_F2 , 16) You should specify these cases in enum vlmul_type get_vlmul (machine_mode mode) And don't change: (define_attr "sew" "" - (cond [(eq_attr "mode" "RVVMF64BI,RVVMF32BI,RVVMF16BI,RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ + (cond [(eq_attr "mode" "RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ RVVM8QI,RVVM4QI,RVVM2QI,RVVM1QI,RVVMF2QI,RVVMF4QI,RVVMF8QI,\ RVVM1x8QI,RVVMF2x8QI,RVVMF4x8QI,RVVMF8x8QI,\ RVVM1x7QI,RVVMF2x7QI,RVVMF4x7QI,RVVMF8x7QI,\ @@ -99,6 +99,18 @@ V1QI,V2QI,V4QI,V8QI,V16QI,V32QI,V64QI,V128QI,V256QI,V512QI,V1024QI,V2048QI,V4096QI,\ V1BI,V2BI,V4BI,V8BI,V16BI,V32BI,V64BI,V128BI,V256BI,V512BI,V1024BI,V2048BI,V4096BI") (const_int 8) + (eq_attr "mode" "RVVMF16BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 16) + (const_int 8)) + (eq_attr "mode" "RVVMF32BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 32) + (const_int 8)) + (eq_attr "mode" "RVVMF64BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 64) + (const_int 8)) (eq_attr "mode" "RVVM8HI,RVVM4HI,RVVM2HI,RVVM1HI,RVVMF2HI,RVVMF4HI,\ RVVM1x8HI,RVVMF2x8HI,RVVMF4x8HI,\ RVVM1x7HI,RVVMF2x7HI,RVVMF4x7HI,\ @@ -159,9 +171,9 @@ (eq_attr "mode" "RVVM4QI,RVVMF2BI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2QI,RVVMF4BI") (symbol_ref "riscv_vector::LMUL_2") (eq_attr "mode" "RVVM1QI,RVVMF8BI") (symbol_ref "riscv_vector::LMUL_1") - (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "riscv_vector::LMUL_F2") - (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "riscv_vector::LMUL_F4") - (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "riscv_vector::LMUL_F8") + (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F2") + (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F4") + (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F8")You should change /* Helper function to get SEW operand. We always have SEW value for all RVV instructions that have VTYPE OP. */ uint8_t get_sew (rtx_insn *rinsn) { return get_attr_sew (rinsn); } /* Helper function to get VLMUL operand. We always have VLMUL value for all RVV instructions that have VTYPE OP. */ enum vlmul_type get_vlmul (rtx_insn *rinsn) { return (enum vlmul_type) get_attr_vlmul (rinsn); } + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE)Why do you add this ? juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 22:11 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Is the patch with !TARGET_XTHEADVECTOR for sext/zext patterns removed OK to commit? https://gcc.gnu.org/pipermail/gcc-patches/2024-January/642657.html ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 18:56 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Yes. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 18:54 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Do you mean removing TARGET_XTHEADVECTOR for sext/zext patterns and then resending the "RISC-V: Handle differences between XTheadvector and Vector" patch? ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:57 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. LGTM. Could you resend the patch "RISC-V: Handle differences between XTheadvector and Vector Thanks. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-11 17:52 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (class th_extract): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (struct th_extract_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): * config/riscv/riscv-vector-builtins.h (enum required_ext): (struct function_group_info): * config/riscv/t-riscv: Add thead-vector-builtins-functions.def * config/riscv/thead-vector.md (@pred_mov_width<vlmem_op_attr><mode>): Add new patterns. (*pred_mov_width<vlmem_op_attr><mode>): Likewise. (@pred_store_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_load_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_store_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_load_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>): (@pred_th_extract<mode>): Likewise. (*pred_th_extract<mode>): Likewise. * config/riscv/thead-vector-builtins-functions.def: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 +++ .../riscv/riscv-vector-builtins-shapes.cc | 160 +++++++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + gcc/config/riscv/riscv-vector-builtins.cc | 70 +++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 250 ++++++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 +++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 +++++ 15 files changed, 1104 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 1aa6e3c6665..b6f6e4ff37e 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2141,6 +2141,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2603,6 +2680,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> vsuxw_obj; +static CONSTEXPR const th_extract vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2894,6 +3002,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (vlb) +BASE (vlh) +BASE (vlw) +BASE (vlbu) +BASE (vlhu) +BASE (vlwu) +BASE (vsb) +BASE (vsh) +BASE (vsw) +BASE (vlsb) +BASE (vlsh) +BASE (vlsw) +BASE (vlsbu) +BASE (vlshu) +BASE (vlswu) +BASE (vssb) +BASE (vssh) +BASE (vssw) +BASE (vlxb) +BASE (vlxh) +BASE (vlxw) +BASE (vlxbu) +BASE (vlxhu) +BASE (vlxwu) +BASE (vsxb) +BASE (vsxh) +BASE (vsxw) +BASE (vsuxb) +BASE (vsuxh) +BASE (vsuxw) +BASE (vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 87c7f43a423..1f2c94d3541 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -280,6 +280,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const vlb; +extern const function_base *const vlh; +extern const function_base *const vlw; +extern const function_base *const vlbu; +extern const function_base *const vlhu; +extern const function_base *const vlwu; +extern const function_base *const vsb; +extern const function_base *const vsh; +extern const function_base *const vsw; +extern const function_base *const vlsb; +extern const function_base *const vlsh; +extern const function_base *const vlsw; +extern const function_base *const vlsbu; +extern const function_base *const vlshu; +extern const function_base *const vlswu; +extern const function_base *const vssb; +extern const function_base *const vssh; +extern const function_base *const vssw; +extern const function_base *const vlxb; +extern const function_base *const vlxh; +extern const function_base *const vlxw; +extern const function_base *const vlxbu; +extern const function_base *const vlxhu; +extern const function_base *const vlxwu; +extern const function_base *const vsxb; +extern const function_base *const vsxh; +extern const function_base *const vsxw; +extern const function_base *const vsuxb; +extern const function_base *const vsuxh; +extern const function_base *const vsuxw; +extern const function_base *const vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..8e90b17a94b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,146 @@ struct indexed_loadstore_def : public function_shape } }; +/* Add one function instance for GROUP, using operand suffix at index OI, + mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ +static void +build_th_loadstore (function_builder &b, const function_group_info &group, + unsigned int pred_idx, unsigned int vec_type_idx) +{ + auto_vec<tree, 5> argument_types; + function_instance function_instance (group.base_name, *group.base, + *group.shape, + group.ops_infos.types[vec_type_idx], + group.preds[pred_idx], &group.ops_infos); + tree return_type = group.ops_infos.ret.get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + b.allocate_argument_types (function_instance, argument_types); + b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + + tree type = builtin_types[group.ops_infos.types[vec_type_idx].index].vector; + if (strstr (group.base_name, "l") + && strstr (group.base_name, "u") + && !TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + if (strstr (group.base_name, "l") + && !strstr (group.base_name, "u") + && TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + machine_mode mode = TYPE_MODE (type); + int sew = GET_MODE_BITSIZE (GET_MODE_INNER (mode)); + if (strstr (group.base_name, "h") && sew == 8) + return; + + if (strstr (group.base_name, "w") && (sew == 8 || sew ==16)) + return; + + b.add_overloaded_function (function_instance, *group.shape); + b.add_unique_function (function_instance, (*group.shape), return_type, + argument_types); +} + +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +772,23 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1251,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1265,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..44b9fec1898 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -934,6 +934,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1481,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2672,38 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * index_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +2882,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..22fed60b4c3 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -123,6 +123,7 @@ enum required_ext ZVKNHB_EXT, /* Crypto vector Zvknhb sub-ext */ ZVKSED_EXT, /* Crypto vector Zvksed sub-ext */ ZVKSH_EXT, /* Crypto vector Zvksh sub-ext */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ }; /* Enumerates the RVV operand types. */ @@ -252,6 +253,8 @@ struct function_group_info return TARGET_ZVKSED; case ZVKSH_EXT: return TARGET_ZVKSH; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; default: gcc_unreachable (); } diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..fd3ba29bae9 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlhu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlwu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vsb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vlsb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlshu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlswu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vssb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vlxb, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxh, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxw, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxbu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxhu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxwu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vsxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..5fe9ba08c4e 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,165 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& riscv_vector::whole_reg_to_reg_move_p (operands, <MODE>mode, 7)" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..3c12c124597 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..30bef369375 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..3c8b5ccc16b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..b7c00404f18 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..17a53012acf --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..b187cfc852b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-12 1:14 ` juzhe.zhong @ 2024-01-12 3:26 ` joshua 0 siblings, 0 replies; 133+ messages in thread From: joshua @ 2024-01-12 3:26 UTC (permalink / raw) To: juzhe.zhong, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 61749 bytes --] I have resent all the patches in a new patchset thread. The TARGET_XTHEADVECTOR for sext/zext in "Handle differences between XTheadvector and Vector" have been removed, and all the changelog have been updated. [PATCH v5] RISC-V: Support XTheadVector extension (gnu.org) <https://gcc.gnu.org/pipermail/gcc-patches/2024-January/642780.html > ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月12日(星期五) 09:14 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: 回复：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. I notice the changelog is incorrect. Plz update the change log: gcc/ChangeLog: * config.gcc: Add files for XTheadVector intrinsics. * config/riscv/autovec.md: Guard XTheadVector. * config/riscv/riscv-c.cc: Add pragma for XTheadVector. * config/riscv/riscv-string.cc (expand_block_move): Guard XTheadVector. * config/riscv/riscv-string.cc (vls_mode_valid_p): Avoid autovec. * config/riscv/riscv-vector-builtins-shapes.cc (check_type): (build_one): New function. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): (DEF_THEAD_RVV_FUNCTION): Add new marcos. (check_required_extensions): (handle_pragma_vector): * config/riscv/riscv-vector-builtins.h (RVV_REQUIRE_VECTOR): (RVV_REQUIRE_XTHEADVECTOR): Add RVV_REQUIRE_VECTOR and RVV_REQUIRE_XTHEADVECTOR. (struct function_group_info): * config/riscv/riscv-vector-switch.def (ENTRY): Disable fractional mode for the XTheadVector extension. (TUPLE_ENTRY): Likewise. * config/riscv/riscv-vsetvl.cc: Add functions for xtheadvector. * config/riscv/riscv.cc (riscv_v_ext_vls_mode_p): Guard XTheadVector. (riscv_v_adjust_bytesize): Likewise. (riscv_preferred_simd_mode): Likewsie. (riscv_autovectorize_vector_modes): Likewise. (riscv_vector_mode_supported_any_target_p): Likewise. (TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P): Likewise. * config/riscv/vector.md: Include thead-vector.md. * config/riscv/riscv_th_vector.h: New file. * config/riscv/thead-vector.md: New file. The changlog doesn't match your patch. I suspect you other patches have same issues. Plz update them and resend. juzhe.zhong@rivai.ai 发件人： joshua <mailto:cooper.joshua@linux.alibaba.com > 发送时间： 2024-01-12 08:49 收件人： 钟居哲 <mailto:juzhe.zhong@rivai.ai >; gcc-patches <mailto:gcc-patches@gcc.gnu.org > 抄送： jim.wilson.gcc <mailto:jim.wilson.gcc@gmail.com >; palmer <mailto:palmer@dabbelt.com >; andrew <mailto:andrew@sifive.com >; philipp.tomsich <mailto:philipp.tomsich@vrull.eu >; Jeff Law <mailto:jeffreyalaw@gmail.com >; Christoph Müllner <mailto:christoph.muellner@vrull.eu >; jinma <mailto:jinma@linux.alibaba.com >; Cooper Qu <mailto:cooper.qu@linux.alibaba.com > 主题： 回复：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. The changes to lmul and sew are only done for vbool, which is RVVMF64BI/RVVMF32BI/RVVMF16BI, in the intrinsics with mask. That is because theadvector has no load/store mask instructions. lmul and sew in other non-bool types do not need to be modified. I don’t think directly change get_sew and get_vlmul is reasonable. + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE)Why do you add this ? theadvector does not have sew in load/storeinstructions. In O2, vector will remove onevsetvl for two consecutive vle with differentsew. However, things will go wrong under thisoptimization for theadvector. ------------------------------------------------------------------ 发件人：钟居哲 <juzhe.zhong@rivai.ai> 发送时间：2024年1月12日(星期五) 07:22 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送："jim.wilson.gcc"<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; Jeff Law<jeffreyalaw@gmail.com>; "Christoph Müllner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; Cooper Qu<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Also, you don't need to change these: -ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) -ENTRY (RVVMF32BI, true, LMUL_F4, 32) -ENTRY (RVVMF16BI, true, LMUL_F2, 16) +ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F8, 64) +ENTRY (RVVMF32BI, true, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F4, 32) +ENTRY (RVVMF16BI, true, TARGET_XTHEADVECTOR ? LMUL_1 : LMUL_F2 , 16) You should specify these cases in enum vlmul_type get_vlmul (machine_mode mode) And don't change: (define_attr "sew" "" - (cond [(eq_attr "mode" "RVVMF64BI,RVVMF32BI,RVVMF16BI,RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ + (cond [(eq_attr "mode" "RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ RVVM8QI,RVVM4QI,RVVM2QI,RVVM1QI,RVVMF2QI,RVVMF4QI,RVVMF8QI,\ RVVM1x8QI,RVVMF2x8QI,RVVMF4x8QI,RVVMF8x8QI,\ RVVM1x7QI,RVVMF2x7QI,RVVMF4x7QI,RVVMF8x7QI,\ @@ -99,6 +99,18 @@ V1QI,V2QI,V4QI,V8QI,V16QI,V32QI,V64QI,V128QI,V256QI,V512QI,V1024QI,V2048QI,V4096QI,\ V1BI,V2BI,V4BI,V8BI,V16BI,V32BI,V64BI,V128BI,V256BI,V512BI,V1024BI,V2048BI,V4096BI") (const_int 8) + (eq_attr "mode" "RVVMF16BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 16) + (const_int 8)) + (eq_attr "mode" "RVVMF32BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 32) + (const_int 8)) + (eq_attr "mode" "RVVMF64BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 64) + (const_int 8)) (eq_attr "mode" "RVVM8HI,RVVM4HI,RVVM2HI,RVVM1HI,RVVMF2HI,RVVMF4HI,\ RVVM1x8HI,RVVMF2x8HI,RVVMF4x8HI,\ RVVM1x7HI,RVVMF2x7HI,RVVMF4x7HI,\ @@ -159,9 +171,9 @@ (eq_attr "mode" "RVVM4QI,RVVMF2BI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2QI,RVVMF4BI") (symbol_ref "riscv_vector::LMUL_2") (eq_attr "mode" "RVVM1QI,RVVMF8BI") (symbol_ref "riscv_vector::LMUL_1") - (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "riscv_vector::LMUL_F2") - (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "riscv_vector::LMUL_F4") - (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "riscv_vector::LMUL_F8") + (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F2") + (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F4") + (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F8")You should change /* Helper function to get SEW operand. We always have SEW value for all RVV instructions that have VTYPE OP. */ uint8_t get_sew (rtx_insn *rinsn) { return get_attr_sew (rinsn); } /* Helper function to get VLMUL operand. We always have VLMUL value for all RVV instructions that have VTYPE OP. */ enum vlmul_type get_vlmul (rtx_insn *rinsn) { return (enum vlmul_type) get_attr_vlmul (rinsn); } + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE)Why do you add this ? juzhe.zhong@rivai.ai 发件人： joshua <mailto:cooper.joshua@linux.alibaba.com > 发送时间： 2024-01-11 22:11 收件人： juzhe.zhong@rivai.ai <mailto:juzhe.zhong@rivai.ai >; gcc-patches <mailto:gcc-patches@gcc.gnu.org > 抄送： Jim Wilson <mailto:jim.wilson.gcc@gmail.com >; palmer <mailto:palmer@dabbelt.com >; andrew <mailto:andrew@sifive.com >; philipp.tomsich <mailto:philipp.tomsich@vrull.eu >; jeffreyalaw <mailto:jeffreyalaw@gmail.com >; christoph.muellner <mailto:christoph.muellner@vrull.eu >; jinma <mailto:jinma@linux.alibaba.com >; cooper.qu <mailto:cooper.qu@linux.alibaba.com > 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Is the patch with !TARGET_XTHEADVECTOR for sext/zext patterns removed OK to commit? https://gcc.gnu.org/pipermail/gcc-patches/2024-January/642657.html <https://gcc.gnu.org/pipermail/gcc-patches/2024-January/642657.html > ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 18:56 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Yes. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-11 18:54 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Do you mean removing TARGET_XTHEADVECTOR for sext/zext patterns and then resending the "RISC-V: Handle differences between XTheadvector and Vector" patch? ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月11日(星期四) 17:57 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. LGTM. Could you resend the patch "RISC-V: Handle differences between XTheadvector and Vector Thanks. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-11 17:52 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (class th_extract): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (struct th_extract_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): * config/riscv/riscv-vector-builtins.h (enum required_ext): (struct function_group_info): * config/riscv/t-riscv: Add thead-vector-builtins-functions.def * config/riscv/thead-vector.md (@pred_mov_width<vlmem_op_attr><mode>): Add new patterns. (*pred_mov_width<vlmem_op_attr><mode>): Likewise. (@pred_store_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_load_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_store_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_load_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>): (@pred_th_extract<mode>): Likewise. (*pred_th_extract<mode>): Likewise. * config/riscv/thead-vector-builtins-functions.def: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 +++ .../riscv/riscv-vector-builtins-shapes.cc | 160 +++++++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + gcc/config/riscv/riscv-vector-builtins.cc | 70 +++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 250 ++++++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 +++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 +++++ 15 files changed, 1104 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 1aa6e3c6665..b6f6e4ff37e 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2141,6 +2141,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2603,6 +2680,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> vsuxw_obj; +static CONSTEXPR const th_extract vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2894,6 +3002,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (vlb) +BASE (vlh) +BASE (vlw) +BASE (vlbu) +BASE (vlhu) +BASE (vlwu) +BASE (vsb) +BASE (vsh) +BASE (vsw) +BASE (vlsb) +BASE (vlsh) +BASE (vlsw) +BASE (vlsbu) +BASE (vlshu) +BASE (vlswu) +BASE (vssb) +BASE (vssh) +BASE (vssw) +BASE (vlxb) +BASE (vlxh) +BASE (vlxw) +BASE (vlxbu) +BASE (vlxhu) +BASE (vlxwu) +BASE (vsxb) +BASE (vsxh) +BASE (vsxw) +BASE (vsuxb) +BASE (vsuxh) +BASE (vsuxw) +BASE (vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 87c7f43a423..1f2c94d3541 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -280,6 +280,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const vlb; +extern const function_base *const vlh; +extern const function_base *const vlw; +extern const function_base *const vlbu; +extern const function_base *const vlhu; +extern const function_base *const vlwu; +extern const function_base *const vsb; +extern const function_base *const vsh; +extern const function_base *const vsw; +extern const function_base *const vlsb; +extern const function_base *const vlsh; +extern const function_base *const vlsw; +extern const function_base *const vlsbu; +extern const function_base *const vlshu; +extern const function_base *const vlswu; +extern const function_base *const vssb; +extern const function_base *const vssh; +extern const function_base *const vssw; +extern const function_base *const vlxb; +extern const function_base *const vlxh; +extern const function_base *const vlxw; +extern const function_base *const vlxbu; +extern const function_base *const vlxhu; +extern const function_base *const vlxwu; +extern const function_base *const vsxb; +extern const function_base *const vsxh; +extern const function_base *const vsxw; +extern const function_base *const vsuxb; +extern const function_base *const vsuxh; +extern const function_base *const vsuxw; +extern const function_base *const vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..8e90b17a94b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,146 @@ struct indexed_loadstore_def : public function_shape } }; +/* Add one function instance for GROUP, using operand suffix at index OI, + mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ +static void +build_th_loadstore (function_builder &b, const function_group_info &group, + unsigned int pred_idx, unsigned int vec_type_idx) +{ + auto_vec<tree, 5> argument_types; + function_instance function_instance (group.base_name, *group.base, + *group.shape, + group.ops_infos.types[vec_type_idx], + group.preds[pred_idx], &group.ops_infos); + tree return_type = group.ops_infos.ret.get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + b.allocate_argument_types (function_instance, argument_types); + b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + + tree type = builtin_types[group.ops_infos.types[vec_type_idx].index].vector; + if (strstr (group.base_name, "l") + && strstr (group.base_name, "u") + && !TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + if (strstr (group.base_name, "l") + && !strstr (group.base_name, "u") + && TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + machine_mode mode = TYPE_MODE (type); + int sew = GET_MODE_BITSIZE (GET_MODE_INNER (mode)); + if (strstr (group.base_name, "h") && sew == 8) + return; + + if (strstr (group.base_name, "w") && (sew == 8 || sew ==16)) + return; + + b.add_overloaded_function (function_instance, *group.shape); + b.add_unique_function (function_instance, (*group.shape), return_type, + argument_types); +} + +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +772,23 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1251,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1265,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..44b9fec1898 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -934,6 +934,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1481,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2672,38 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * index_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +2882,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..22fed60b4c3 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -123,6 +123,7 @@ enum required_ext ZVKNHB_EXT, /* Crypto vector Zvknhb sub-ext */ ZVKSED_EXT, /* Crypto vector Zvksed sub-ext */ ZVKSH_EXT, /* Crypto vector Zvksh sub-ext */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ }; /* Enumerates the RVV operand types. */ @@ -252,6 +253,8 @@ struct function_group_info return TARGET_ZVKSED; case ZVKSH_EXT: return TARGET_ZVKSH; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; default: gcc_unreachable (); } diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..fd3ba29bae9 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlhu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlwu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vsb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vlsb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlshu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlswu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vssb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vlxb, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxh, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxw, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxbu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxhu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxwu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vsxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..5fe9ba08c4e 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,165 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& riscv_vector::whole_reg_to_reg_move_p (operands, <MODE>mode, 7)" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..3c12c124597 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..30bef369375 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..3c8b5ccc16b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..b7c00404f18 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..17a53012acf --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..b187cfc852b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* [PATCH v4] RISC-V: Fix register overlap issue for some xtheadvector instructions 2023-12-29 4:03 [PATCH v4] RISC-V: Support XTheadVector extension Jun Sha (Joshua) ` (5 preceding siblings ...) 2023-12-29 4:21 ` [PATCH v4 6/6] RISC-V: Add support for xtheadvector-specific intrinsics Jun Sha (Joshua) @ 2024-01-03 2:37 ` Jun Sha (Joshua) 2024-01-03 7:54 ` Jun Sha (Joshua) 2024-01-03 2:39 ` [PATCH v4] RISC-V: Rewrite some instructions using ASM targethook Jun Sha (Joshua) 2024-01-12 3:18 ` [PATCH v5] RISC-V: Support XTheadVector extension Jun Sha (Joshua) 8 siblings, 1 reply; 133+ messages in thread From: Jun Sha (Joshua) @ 2024-01-03 2:37 UTC (permalink / raw) To: gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, Jun Sha (Joshua), Jin Ma, Xianmiao Qu For th.vmadc/th.vmsbc as well as narrowing arithmetic instructions and floating-point compare instructions, an illegal instruction exception will be raised if the destination vector register overlaps a source vector register group. To handle this issue, we use "group_overlap" and "enabled" attribute to disable some alternatives for xtheadvector. gcc/ChangeLog: * config/riscv/riscv.md (none,W21,W42,W84,W43,W86,W87,W0): (none,W21,W42,W84,W43,W86,W87,W0,th): * config/riscv/vector.md: Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config/riscv/riscv.md | 6 +- gcc/config/riscv/vector.md | 314 +++++++++++++++++++++---------------- 2 files changed, 185 insertions(+), 135 deletions(-) diff --git a/gcc/config/riscv/riscv.md b/gcc/config/riscv/riscv.md index 68f7203b676..d736501784d 100644 --- a/gcc/config/riscv/riscv.md +++ b/gcc/config/riscv/riscv.md @@ -504,7 +504,7 @@ ;; Widening instructions have group-overlap constraints. Those are only ;; valid for certain register-group sizes. This attribute marks the ;; alternatives not matching the required register-group size as disabled. -(define_attr "group_overlap" "none,W21,W42,W84,W43,W86,W87,W0" +(define_attr "group_overlap" "none,W21,W42,W84,W43,W86,W87,W0,th" (const_string "none")) (define_attr "group_overlap_valid" "no,yes" @@ -543,6 +543,10 @@ (and (eq_attr "group_overlap" "W0") (match_test "riscv_get_v_regno_alignment (GET_MODE (operands[0])) > 1")) (const_string "no") + + (and (eq_attr "group_overlap" "th") + (match_test "TARGET_XTHEADVECTOR")) + (const_string "no") ] (const_string "yes"))) diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index 5fa30716143..77eaba16c97 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -3255,7 +3255,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "th,none,none")]) (define_insn "@pred_msbc<mode>" [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr") @@ -3274,7 +3275,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "th,th,none")]) (define_insn "@pred_madc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3294,7 +3296,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "th,none")]) (define_insn "@pred_msbc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3314,7 +3317,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "th,none")]) (define_expand "@pred_madc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3363,7 +3367,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "th,none")]) (define_insn "*pred_madc<mode>_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3384,7 +3389,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "th,none")]) (define_expand "@pred_msbc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3433,7 +3439,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "th,none")]) (define_insn "*pred_msbc<mode>_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3454,7 +3461,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "th,none")]) (define_insn "@pred_madc<mode>_overflow" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr, &vr") @@ -3472,7 +3480,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "th,none,none")]) (define_insn "@pred_msbc<mode>_overflow" [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") @@ -3490,7 +3499,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "th,th,none,none")]) (define_insn "@pred_madc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3509,7 +3519,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "th,none")]) (define_insn "@pred_msbc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3528,7 +3539,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "th,none")]) (define_expand "@pred_madc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3575,7 +3587,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "th,none")]) (define_insn "*pred_madc<mode>_overflow_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3595,7 +3608,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "th,none")]) (define_expand "@pred_msbc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3642,7 +3656,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "th,none")]) (define_insn "*pred_msbc<mode>_overflow_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3662,7 +3677,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "th,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated integer unary operations @@ -3982,7 +3998,8 @@ "TARGET_VECTOR" "vn<insn>.w%o4\t%0,%3,%v4%p1" [(set_attr "type" "vnshift") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "group_overlap" "none,none,th,th,none,th,none,none,none,th,none,none")]) (define_insn "@pred_narrow_<optab><mode>_scalar" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -4003,7 +4020,8 @@ "TARGET_VECTOR" "vn<insn>.w%o4\t%0,%3,%4%p1" [(set_attr "type" "vnshift") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "group_overlap" "none,none,th,th,none,none")]) ;; vncvt.x.x.w (define_insn "@pred_trunc<mode>" @@ -4027,7 +4045,8 @@ (set_attr "vl_op_idx" "4") (set (attr "ta") (symbol_ref "riscv_vector::get_ta(operands[5])")) (set (attr "ma") (symbol_ref "riscv_vector::get_ma(operands[6])")) - (set (attr "avl_type_idx") (const_int 7))]) + (set (attr "avl_type_idx") (const_int 7)) + (set_attr "group_overlap" "none,none,th,th,none,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated fixed-point operations @@ -4433,7 +4452,8 @@ "TARGET_VECTOR" "vnclip<v_su>.w%o4\t%0,%3,%v4%p1" [(set_attr "type" "vnclip") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "group_overlap" "th,th,th,th,th,th,none,none,th,th,none,none")]) (define_insn "@pred_narrow_clip<v_su><mode>_scalar" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -4455,7 +4475,8 @@ "TARGET_VECTOR" "vnclip<v_su>.w%o4\t%0,%3,%4%p1" [(set_attr "type" "vnclip") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "group_overlap" "th,th,th,th,none,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated integer comparison operations @@ -4506,23 +4527,24 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr, &vr, &vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") - (match_operand 7 "const_int_operand" " i, i, i, i") - (match_operand 8 "const_int_operand" " i, i, i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK, rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i, i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_ltge_operator" - [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr") - (match_operand:V_VLSI 5 "vector_arith_operand" " vr, vr, vi, vi")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] + [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr, vr, vr, vr, vr") + (match_operand:V_VLSI 5 "vector_arith_operand" " vr, vr, vi, vi, vr, vr, vi, vi")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0, vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,th,th,none,none,none,none")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_narrow" @@ -4542,7 +4564,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,th,th,th,th,none,none")]) (define_expand "@pred_ltge<mode>" [(set (match_operand:<VM> 0 "register_operand") @@ -4586,23 +4609,24 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_ltge<mode>" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") - (match_operand 7 "const_int_operand" " i, i, i, i") - (match_operand 8 "const_int_operand" " i, i, i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "ltge_operator" - [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr") - (match_operand:V_VLSI 5 "vector_neg_arith_operand" " vr, vr, vj, vj")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] + [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr, vr, vr") + (match_operand:V_VLSI 5 "vector_neg_arith_operand" " vr, vr, vj, vj, vr, vj")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0, vu, vu")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,th,th,none,none")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_ltge<mode>_narrow" @@ -4622,7 +4646,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,th,th,th,th,none,none")]) (define_expand "@pred_cmp<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -4668,24 +4693,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_eqge_operator" - [(match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr") + [(match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr, vr") (vec_duplicate:V_VLSI_QHS - (match_operand:<VEL> 5 "register_operand" " r, r"))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r"))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,none")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_scalar_narrow" @@ -4706,7 +4732,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,none,none")]) (define_expand "@pred_eqne<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -4752,24 +4779,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSI_QHS - (match_operand:<VEL> 5 "register_operand" " r, r")) - (match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r")) + (match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,none")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_eqne<mode>_scalar_narrow" @@ -4790,7 +4818,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,none,none")]) ;; Handle GET_MODE_INNER (mode) = DImode. We need to split them since ;; we need to deal with SEW = 64 in RV32 system. @@ -4917,24 +4946,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_eqge_operator" - [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr") + [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr") (vec_duplicate:V_VLSI_D - (match_operand:<VEL> 5 "register_operand" " r, r"))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r"))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,none")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_scalar_narrow" @@ -4955,28 +4985,30 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,none,none")]) ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSI_D - (match_operand:<VEL> 5 "register_operand" " r, r")) - (match_operand:V_VLSI_D 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r")) + (match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,none")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_eqne<mode>_scalar_narrow" @@ -4997,7 +5029,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,none,none")]) (define_insn "*pred_cmp<mode>_extended_scalar_merge_tie_mask" [(set (match_operand:<VM> 0 "register_operand" "=vm") @@ -5026,25 +5059,26 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_extended_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_eqge_operator" - [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr") + [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr") (vec_duplicate:V_VLSI_D (sign_extend:<VEL> - (match_operand:<VSUBEL> 5 "register_operand" " r, r")))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VSUBEL> 5 "register_operand" " r, r, r")))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,none")]) (define_insn "*pred_cmp<mode>_extended_scalar_narrow" [(set (match_operand:<VM> 0 "register_operand" "=vm, vr, vr, &vr, &vr") @@ -5065,7 +5099,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,none,none")]) (define_insn "*pred_eqne<mode>_extended_scalar_merge_tie_mask" [(set (match_operand:<VM> 0 "register_operand" "=vm") @@ -5094,25 +5129,26 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_extended_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSI_D (sign_extend:<VEL> - (match_operand:<VSUBEL> 5 "register_operand" " r, r"))) - (match_operand:V_VLSI_D 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VSUBEL> 5 "register_operand" " r, r, r"))) + (match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,none")]) (define_insn "*pred_eqne<mode>_extended_scalar_narrow" [(set (match_operand:<VM> 0 "register_operand" "=vm, vr, vr, &vr, &vr") @@ -5133,7 +5169,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,none,none")]) ;; GE, vmsge.vx/vmsgeu.vx ;; @@ -7322,23 +7359,24 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "signed_order_operator" - [(match_operand:V_VLSF 4 "register_operand" " vr, vr") - (match_operand:V_VLSF 5 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + [(match_operand:V_VLSF 4 "register_operand" " vr, vr, vr, vr") + (match_operand:V_VLSF 5 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vmf%B3.vv\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,none,none")]) (define_insn "*pred_cmp<mode>_narrow_merge_tie_mask" [(set (match_operand:<VM> 0 "register_operand" "=vm") @@ -7381,7 +7419,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vmf%B3.vv\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,th,th,th,th,none,none")]) (define_expand "@pred_cmp<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -7427,24 +7466,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "signed_order_operator" - [(match_operand:V_VLSF 4 "register_operand" " vr, vr") + [(match_operand:V_VLSF 4 "register_operand" " vr, vr, vr") (vec_duplicate:V_VLSF - (match_operand:<VEL> 5 "register_operand" " f, f"))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " f, f, f"))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,none")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_scalar_narrow" @@ -7465,7 +7505,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,none,none")]) (define_expand "@pred_eqne<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -7511,24 +7552,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSF - (match_operand:<VEL> 5 "register_operand" " f, f")) - (match_operand:V_VLSF 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " f, f, f")) + (match_operand:V_VLSF 4 "register_operand" " vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,none")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_eqne<mode>_scalar_narrow" @@ -7549,7 +7591,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,none,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated floating-point merge @@ -7769,7 +7812,8 @@ [(set_attr "type" "vfncvtftoi") (set_attr "mode" "<VNCONVERT>") (set (attr "frm_mode") - (symbol_ref "riscv_vector::get_frm_mode (operands[8])"))]) + (symbol_ref "riscv_vector::get_frm_mode (operands[8])")) + (set_attr "group_overlap" "none,none,th,th,none,none")]) (define_insn "@pred_narrow_<fix_cvt><mode>" [(set (match_operand:<VNCONVERT> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -7811,7 +7855,8 @@ [(set_attr "type" "vfncvtitof") (set_attr "mode" "<VNCONVERT>") (set (attr "frm_mode") - (symbol_ref "riscv_vector::get_frm_mode (operands[8])"))]) + (symbol_ref "riscv_vector::get_frm_mode (operands[8])")) + (set_attr "group_overlap" "none,none,th,th,none,none")]) (define_insn "@pred_trunc<mode>" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -7834,7 +7879,8 @@ [(set_attr "type" "vfncvtftof") (set_attr "mode" "<V_DOUBLE_TRUNC>") (set (attr "frm_mode") - (symbol_ref "riscv_vector::get_frm_mode (operands[8])"))]) + (symbol_ref "riscv_vector::get_frm_mode (operands[8])")) + (set_attr "group_overlap" "none,none,th,th,none,none")]) (define_insn "@pred_rod_trunc<mode>" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* [PATCH v4] RISC-V: Fix register overlap issue for some xtheadvector instructions 2024-01-03 2:37 ` [PATCH v4] RISC-V: Fix register overlap issue for some xtheadvector instructions Jun Sha (Joshua) @ 2024-01-03 7:54 ` Jun Sha (Joshua) 2024-01-10 6:02 ` [PATCH v5] " Jun Sha (Joshua) 2024-01-10 6:51 ` Jun Sha (Joshua) 0 siblings, 2 replies; 133+ messages in thread From: Jun Sha (Joshua) @ 2024-01-03 7:54 UTC (permalink / raw) To: gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, Jun Sha (Joshua), Jin Ma, Xianmiao Qu For th.vmadc/th.vmsbc as well as narrowing arithmetic instructions and floating-point compare instructions, an illegal instruction exception will be raised if the destination vector register overlaps a source vector register group. To handle this issue, we use "group_overlap" and "enabled" attribute to disable some alternatives for xtheadvector. gcc/ChangeLog: * config/riscv/riscv.md (none,W21,W42,W84,W43,W86,W87,W0): (none,W21,W42,W84,W43,W86,W87,W0,th): Add group-overlap constraint for xtheadvector. * config/riscv/vector.md: Disable alternatives that destination register overlaps source register group for xtheadvector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config/riscv/riscv.md | 6 +- gcc/config/riscv/vector.md | 314 +++++++++++++++++++++---------------- 2 files changed, 185 insertions(+), 135 deletions(-) diff --git a/gcc/config/riscv/riscv.md b/gcc/config/riscv/riscv.md index 68f7203b676..d736501784d 100644 --- a/gcc/config/riscv/riscv.md +++ b/gcc/config/riscv/riscv.md @@ -504,7 +504,7 @@ ;; Widening instructions have group-overlap constraints. Those are only ;; valid for certain register-group sizes. This attribute marks the ;; alternatives not matching the required register-group size as disabled. -(define_attr "group_overlap" "none,W21,W42,W84,W43,W86,W87,W0" +(define_attr "group_overlap" "none,W21,W42,W84,W43,W86,W87,W0,th" (const_string "none")) (define_attr "group_overlap_valid" "no,yes" @@ -543,6 +543,10 @@ (and (eq_attr "group_overlap" "W0") (match_test "riscv_get_v_regno_alignment (GET_MODE (operands[0])) > 1")) (const_string "no") + + (and (eq_attr "group_overlap" "th") + (match_test "TARGET_XTHEADVECTOR")) + (const_string "no") ] (const_string "yes"))) diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index cb30c9ae97c..63d0573d4aa 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -3248,7 +3248,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "th,none,none")]) (define_insn "@pred_msbc<mode>" [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr") @@ -3267,7 +3268,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "th,th,none")]) (define_insn "@pred_madc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3287,7 +3289,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "th,none")]) (define_insn "@pred_msbc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3307,7 +3310,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "th,none")]) (define_expand "@pred_madc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3356,7 +3360,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "th,none")]) (define_insn "*pred_madc<mode>_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3377,7 +3382,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "th,none")]) (define_expand "@pred_msbc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3426,7 +3432,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "th,none")]) (define_insn "*pred_msbc<mode>_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3447,7 +3454,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "th,none")]) (define_insn "@pred_madc<mode>_overflow" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr, &vr") @@ -3465,7 +3473,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "th,none,none")]) (define_insn "@pred_msbc<mode>_overflow" [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") @@ -3483,7 +3492,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "th,th,none,none")]) (define_insn "@pred_madc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3502,7 +3512,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "th,none")]) (define_insn "@pred_msbc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3521,7 +3532,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "th,none")]) (define_expand "@pred_madc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3568,7 +3580,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "th,none")]) (define_insn "*pred_madc<mode>_overflow_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3588,7 +3601,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "th,none")]) (define_expand "@pred_msbc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3635,7 +3649,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "th,none")]) (define_insn "*pred_msbc<mode>_overflow_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3655,7 +3670,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "th,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated integer unary operations @@ -3975,7 +3991,8 @@ "TARGET_VECTOR" "vn<insn>.w%o4\t%0,%3,%v4%p1" [(set_attr "type" "vnshift") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "group_overlap" "none,none,th,th,none,th,none,none,none,th,none,none")]) (define_insn "@pred_narrow_<optab><mode>_scalar" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -3996,7 +4013,8 @@ "TARGET_VECTOR" "vn<insn>.w%o4\t%0,%3,%4%p1" [(set_attr "type" "vnshift") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "group_overlap" "none,none,th,th,none,none")]) ;; vncvt.x.x.w (define_insn "@pred_trunc<mode>" @@ -4020,7 +4038,8 @@ (set_attr "vl_op_idx" "4") (set (attr "ta") (symbol_ref "riscv_vector::get_ta(operands[5])")) (set (attr "ma") (symbol_ref "riscv_vector::get_ma(operands[6])")) - (set (attr "avl_type_idx") (const_int 7))]) + (set (attr "avl_type_idx") (const_int 7)) + (set_attr "group_overlap" "none,none,th,th,none,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated fixed-point operations @@ -4426,7 +4445,8 @@ "TARGET_VECTOR" "vnclip<v_su>.w%o4\t%0,%3,%v4%p1" [(set_attr "type" "vnclip") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "group_overlap" "th,th,th,th,th,th,none,none,th,th,none,none")]) (define_insn "@pred_narrow_clip<v_su><mode>_scalar" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -4448,7 +4468,8 @@ "TARGET_VECTOR" "vnclip<v_su>.w%o4\t%0,%3,%4%p1" [(set_attr "type" "vnclip") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "group_overlap" "th,th,th,th,none,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated integer comparison operations @@ -4499,23 +4520,24 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr, &vr, &vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") - (match_operand 7 "const_int_operand" " i, i, i, i") - (match_operand 8 "const_int_operand" " i, i, i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK, rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i, i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_ltge_operator" - [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr") - (match_operand:V_VLSI 5 "vector_arith_operand" " vr, vr, vi, vi")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] + [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr, vr, vr, vr, vr") + (match_operand:V_VLSI 5 "vector_arith_operand" " vr, vr, vi, vi, vr, vr, vi, vi")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0, vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,th,th,none,none,none,none")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_narrow" @@ -4535,7 +4557,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,th,th,th,th,none,none")]) (define_expand "@pred_ltge<mode>" [(set (match_operand:<VM> 0 "register_operand") @@ -4579,23 +4602,24 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_ltge<mode>" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr, &vr, &vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") - (match_operand 7 "const_int_operand" " i, i, i, i") - (match_operand 8 "const_int_operand" " i, i, i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK, rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i, i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "ltge_operator" - [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr") - (match_operand:V_VLSI 5 "vector_neg_arith_operand" " vr, vr, vj, vj")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] + [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr, vr, vr, vr, vr") + (match_operand:V_VLSI 5 "vector_neg_arith_operand" " vr, vr, vj, vj, vr, vr, vj, vj")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0, vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,th,th,none,none,none,none")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_ltge<mode>_narrow" @@ -4615,7 +4639,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,th,th,th,th,none,none")]) (define_expand "@pred_cmp<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -4661,24 +4686,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_eqge_operator" - [(match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr") + [(match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSI_QHS - (match_operand:<VEL> 5 "register_operand" " r, r"))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r"))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,none,none")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_scalar_narrow" @@ -4699,7 +4725,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,none,none")]) (define_expand "@pred_eqne<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -4745,24 +4772,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSI_QHS - (match_operand:<VEL> 5 "register_operand" " r, r")) - (match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r")) + (match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,none,none")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_eqne<mode>_scalar_narrow" @@ -4783,7 +4811,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,none,none")]) ;; Handle GET_MODE_INNER (mode) = DImode. We need to split them since ;; we need to deal with SEW = 64 in RV32 system. @@ -4910,24 +4939,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_eqge_operator" - [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr") + [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSI_D - (match_operand:<VEL> 5 "register_operand" " r, r"))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r"))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,none,none")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_scalar_narrow" @@ -4948,28 +4978,30 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,none,none")]) ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSI_D - (match_operand:<VEL> 5 "register_operand" " r, r")) - (match_operand:V_VLSI_D 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r")) + (match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,none,none")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_eqne<mode>_scalar_narrow" @@ -4990,7 +5022,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,none,none")]) (define_insn "*pred_cmp<mode>_extended_scalar_merge_tie_mask" [(set (match_operand:<VM> 0 "register_operand" "=vm") @@ -5019,25 +5052,26 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_extended_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_eqge_operator" - [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr") + [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSI_D (sign_extend:<VEL> - (match_operand:<VSUBEL> 5 "register_operand" " r, r")))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VSUBEL> 5 "register_operand" " r, r, r, r")))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,none,none")]) (define_insn "*pred_cmp<mode>_extended_scalar_narrow" [(set (match_operand:<VM> 0 "register_operand" "=vm, vr, vr, &vr, &vr") @@ -5058,7 +5092,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,none,none")]) (define_insn "*pred_eqne<mode>_extended_scalar_merge_tie_mask" [(set (match_operand:<VM> 0 "register_operand" "=vm") @@ -5087,25 +5122,26 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_extended_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSI_D (sign_extend:<VEL> - (match_operand:<VSUBEL> 5 "register_operand" " r, r"))) - (match_operand:V_VLSI_D 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VSUBEL> 5 "register_operand" " r, r, r, r"))) + (match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,none,none")]) (define_insn "*pred_eqne<mode>_extended_scalar_narrow" [(set (match_operand:<VM> 0 "register_operand" "=vm, vr, vr, &vr, &vr") @@ -5126,7 +5162,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,none,none")]) ;; GE, vmsge.vx/vmsgeu.vx ;; @@ -7315,23 +7352,24 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "signed_order_operator" - [(match_operand:V_VLSF 4 "register_operand" " vr, vr") - (match_operand:V_VLSF 5 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + [(match_operand:V_VLSF 4 "register_operand" " vr, vr, vr, vr") + (match_operand:V_VLSF 5 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vmf%B3.vv\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,none,none")]) (define_insn "*pred_cmp<mode>_narrow_merge_tie_mask" [(set (match_operand:<VM> 0 "register_operand" "=vm") @@ -7374,7 +7412,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vmf%B3.vv\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,th,th,th,th,none,none")]) (define_expand "@pred_cmp<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -7420,24 +7459,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "signed_order_operator" - [(match_operand:V_VLSF 4 "register_operand" " vr, vr") + [(match_operand:V_VLSF 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSF - (match_operand:<VEL> 5 "register_operand" " f, f"))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " f, f, f, f"))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,none,none")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_scalar_narrow" @@ -7458,7 +7498,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,none,none")]) (define_expand "@pred_eqne<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -7504,24 +7545,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSF - (match_operand:<VEL> 5 "register_operand" " f, f")) - (match_operand:V_VLSF 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " f, f, f, f")) + (match_operand:V_VLSF 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,none,none")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_eqne<mode>_scalar_narrow" @@ -7542,7 +7584,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,none,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated floating-point merge @@ -7762,7 +7805,8 @@ [(set_attr "type" "vfncvtftoi") (set_attr "mode" "<VNCONVERT>") (set (attr "frm_mode") - (symbol_ref "riscv_vector::get_frm_mode (operands[8])"))]) + (symbol_ref "riscv_vector::get_frm_mode (operands[8])")) + (set_attr "group_overlap" "none,none,th,th,none,none")]) (define_insn "@pred_narrow_<fix_cvt><mode>" [(set (match_operand:<VNCONVERT> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -7804,7 +7848,8 @@ [(set_attr "type" "vfncvtitof") (set_attr "mode" "<VNCONVERT>") (set (attr "frm_mode") - (symbol_ref "riscv_vector::get_frm_mode (operands[8])"))]) + (symbol_ref "riscv_vector::get_frm_mode (operands[8])")) + (set_attr "group_overlap" "none,none,th,th,none,none")]) (define_insn "@pred_trunc<mode>" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -7827,7 +7872,8 @@ [(set_attr "type" "vfncvtftof") (set_attr "mode" "<V_DOUBLE_TRUNC>") (set (attr "frm_mode") - (symbol_ref "riscv_vector::get_frm_mode (operands[8])"))]) + (symbol_ref "riscv_vector::get_frm_mode (operands[8])")) + (set_attr "group_overlap" "none,none,th,th,none,none")]) (define_insn "@pred_rod_trunc<mode>" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* [PATCH v5] RISC-V: Fix register overlap issue for some xtheadvector instructions 2024-01-03 7:54 ` Jun Sha (Joshua) @ 2024-01-10 6:02 ` Jun Sha (Joshua) 2024-01-10 6:37 ` juzhe.zhong 2024-01-10 6:51 ` Jun Sha (Joshua) 1 sibling, 1 reply; 133+ messages in thread From: Jun Sha (Joshua) @ 2024-01-10 6:02 UTC (permalink / raw) To: gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, Jun Sha (Joshua), Jin Ma, Xianmiao Qu For th.vmadc/th.vmsbc as well as narrowing arithmetic instructions and floating-point compare instructions, an illegal instruction exception will be raised if the destination vector register overlaps a source vector register group. To handle this issue, we use "group_overlap" and "enabled" attribute to disable some alternatives for xtheadvector. gcc/ChangeLog: * config/riscv/riscv.md (none,W21,W42,W84,W43,W86,W87,W0): (none,W21,W42,W84,W43,W86,W87,W0,th): Add group-overlap constraint for xtheadvector. * config/riscv/vector.md: Disable alternatives that destination register overlaps source register group for xtheadvector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config/riscv/riscv.md | 12 +- gcc/config/riscv/vector.md | 314 +++++++++++++++++++++---------------- 2 files changed, 190 insertions(+), 136 deletions(-) diff --git a/gcc/config/riscv/riscv.md b/gcc/config/riscv/riscv.md index 84212430dc0..2fe15fd7340 100644 --- a/gcc/config/riscv/riscv.md +++ b/gcc/config/riscv/riscv.md @@ -537,7 +537,7 @@ ;; Widening instructions have group-overlap constraints. Those are only ;; valid for certain register-group sizes. This attribute marks the ;; alternatives not matching the required register-group size as disabled. -(define_attr "group_overlap" "none,W21,W42,W84,W43,W86,W87,W0" +(define_attr "group_overlap" "none,W21,W42,W84,W43,W86,W87,W0,th,rvv" (const_string "none")) (define_attr "group_overlap_valid" "no,yes" @@ -576,7 +576,15 @@ (and (eq_attr "group_overlap" "W0") (match_test "riscv_get_v_regno_alignment (GET_MODE (operands[0])) > 1")) (const_string "no") - ] + + (and (eq_attr "group_overlap" "th") + (match_test "TARGET_XTHEADVECTOR")) + (const_string "no") + + (and (eq_attr "group_overlap" "rvv") + (match_test "TARGET_VECTOR && !TARGET_XTHEADVECTOR")) + (const_string "no") + ] (const_string "yes"))) ;; Attribute to control enable or disable instructions. diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index 3eb6daafbc2..cd83c1f3321 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -3260,7 +3260,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "th,none,none")]) (define_insn "@pred_msbc<mode>" [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr") @@ -3279,7 +3280,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "th,th,none")]) (define_insn "@pred_madc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3299,7 +3301,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "th,none")]) (define_insn "@pred_msbc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3319,7 +3322,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "th,none")]) (define_expand "@pred_madc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3368,7 +3372,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "th,none")]) (define_insn "*pred_madc<mode>_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3389,7 +3394,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "th,none")]) (define_expand "@pred_msbc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3438,7 +3444,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "th,none")]) (define_insn "*pred_msbc<mode>_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3459,7 +3466,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "th,none")]) (define_insn "@pred_madc<mode>_overflow" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr, &vr") @@ -3477,7 +3485,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "th,none,none")]) (define_insn "@pred_msbc<mode>_overflow" [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") @@ -3495,7 +3504,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "th,th,none,none")]) (define_insn "@pred_madc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3514,7 +3524,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "th,none")]) (define_insn "@pred_msbc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3533,7 +3544,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "th,none")]) (define_expand "@pred_madc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3580,7 +3592,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "th,none")]) (define_insn "*pred_madc<mode>_overflow_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3600,7 +3613,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "th,none")]) (define_expand "@pred_msbc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3647,7 +3661,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "th,none")]) (define_insn "*pred_msbc<mode>_overflow_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3667,7 +3682,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "th,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated integer unary operations @@ -3987,7 +4003,8 @@ "TARGET_VECTOR" "vn<insn>.w%o4\t%0,%3,%v4%p1" [(set_attr "type" "vnshift") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "group_overlap" "none,none,th,th,none,th,none,none,none,th,none,none")]) (define_insn "@pred_narrow_<optab><mode>_scalar" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -4008,7 +4025,8 @@ "TARGET_VECTOR" "vn<insn>.w%o4\t%0,%3,%4%p1" [(set_attr "type" "vnshift") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "group_overlap" "none,none,th,th,none,none")]) ;; vncvt.x.x.w (define_insn "@pred_trunc<mode>" @@ -4032,7 +4050,8 @@ (set_attr "vl_op_idx" "4") (set (attr "ta") (symbol_ref "riscv_vector::get_ta(operands[5])")) (set (attr "ma") (symbol_ref "riscv_vector::get_ma(operands[6])")) - (set (attr "avl_type_idx") (const_int 7))]) + (set (attr "avl_type_idx") (const_int 7)) + (set_attr "group_overlap" "none,none,th,th,none,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated fixed-point operations @@ -4438,7 +4457,8 @@ "TARGET_VECTOR" "vnclip<v_su>.w%o4\t%0,%3,%v4%p1" [(set_attr "type" "vnclip") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "group_overlap" "th,th,th,th,th,th,none,none,th,th,none,none")]) (define_insn "@pred_narrow_clip<v_su><mode>_scalar" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -4460,7 +4480,8 @@ "TARGET_VECTOR" "vnclip<v_su>.w%o4\t%0,%3,%4%p1" [(set_attr "type" "vnclip") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "group_overlap" "th,th,th,th,none,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated integer comparison operations @@ -4511,23 +4532,24 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr, &vr, &vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") - (match_operand 7 "const_int_operand" " i, i, i, i") - (match_operand 8 "const_int_operand" " i, i, i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK, rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i, i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_ltge_operator" - [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr") - (match_operand:V_VLSI 5 "vector_arith_operand" " vr, vr, vi, vi")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] + [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr, vr, vr, vr, vr") + (match_operand:V_VLSI 5 "vector_arith_operand" " vr, vr, vi, vi, vr, vr, vi, vi")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0, vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,th,th,rvv,rvv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_narrow" @@ -4547,7 +4569,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,th,th,th,th,none,none")]) (define_expand "@pred_ltge<mode>" [(set (match_operand:<VM> 0 "register_operand") @@ -4591,23 +4614,24 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_ltge<mode>" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr, &vr, &vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") - (match_operand 7 "const_int_operand" " i, i, i, i") - (match_operand 8 "const_int_operand" " i, i, i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK, rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i, i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "ltge_operator" - [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr") - (match_operand:V_VLSI 5 "vector_neg_arith_operand" " vr, vr, vj, vj")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] + [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr, vr, vr, vr, vr") + (match_operand:V_VLSI 5 "vector_neg_arith_operand" " vr, vr, vj, vj, vr, vr, vj, vj")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0, vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,th,th,rvv,rvv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_ltge<mode>_narrow" @@ -4627,7 +4651,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,th,th,th,th,none,none")]) (define_expand "@pred_cmp<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -4673,24 +4698,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_eqge_operator" - [(match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr") + [(match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSI_QHS - (match_operand:<VEL> 5 "register_operand" " r, r"))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r"))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_scalar_narrow" @@ -4711,7 +4737,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,none,none")]) (define_expand "@pred_eqne<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -4757,24 +4784,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSI_QHS - (match_operand:<VEL> 5 "register_operand" " r, r")) - (match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r")) + (match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_eqne<mode>_scalar_narrow" @@ -4795,7 +4823,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,none,none")]) ;; Handle GET_MODE_INNER (mode) = DImode. We need to split them since ;; we need to deal with SEW = 64 in RV32 system. @@ -4922,24 +4951,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_eqge_operator" - [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr") + [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSI_D - (match_operand:<VEL> 5 "register_operand" " r, r"))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r"))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_scalar_narrow" @@ -4960,28 +4990,30 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,none,none")]) ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSI_D - (match_operand:<VEL> 5 "register_operand" " r, r")) - (match_operand:V_VLSI_D 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r")) + (match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_eqne<mode>_scalar_narrow" @@ -5002,7 +5034,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,none,none")]) (define_insn "*pred_cmp<mode>_extended_scalar_merge_tie_mask" [(set (match_operand:<VM> 0 "register_operand" "=vm") @@ -5031,25 +5064,26 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_extended_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_eqge_operator" - [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr") + [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSI_D (sign_extend:<VEL> - (match_operand:<VSUBEL> 5 "register_operand" " r, r")))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VSUBEL> 5 "register_operand" " r, r, r, r")))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,rvv,rvv")]) (define_insn "*pred_cmp<mode>_extended_scalar_narrow" [(set (match_operand:<VM> 0 "register_operand" "=vm, vr, vr, &vr, &vr") @@ -5070,7 +5104,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,none,none")]) (define_insn "*pred_eqne<mode>_extended_scalar_merge_tie_mask" [(set (match_operand:<VM> 0 "register_operand" "=vm") @@ -5099,25 +5134,26 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_extended_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSI_D (sign_extend:<VEL> - (match_operand:<VSUBEL> 5 "register_operand" " r, r"))) - (match_operand:V_VLSI_D 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VSUBEL> 5 "register_operand" " r, r, r, r"))) + (match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,rvv,rvv")]) (define_insn "*pred_eqne<mode>_extended_scalar_narrow" [(set (match_operand:<VM> 0 "register_operand" "=vm, vr, vr, &vr, &vr") @@ -5138,7 +5174,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,none,none")]) ;; GE, vmsge.vx/vmsgeu.vx ;; @@ -7327,23 +7364,24 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "signed_order_operator" - [(match_operand:V_VLSF 4 "register_operand" " vr, vr") - (match_operand:V_VLSF 5 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + [(match_operand:V_VLSF 4 "register_operand" " vr, vr, vr, vr") + (match_operand:V_VLSF 5 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vmf%B3.vv\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,rvv,rvv")]) (define_insn "*pred_cmp<mode>_narrow_merge_tie_mask" [(set (match_operand:<VM> 0 "register_operand" "=vm") @@ -7386,7 +7424,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vmf%B3.vv\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,th,th,th,th,none,none")]) (define_expand "@pred_cmp<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -7432,24 +7471,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "signed_order_operator" - [(match_operand:V_VLSF 4 "register_operand" " vr, vr") + [(match_operand:V_VLSF 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSF - (match_operand:<VEL> 5 "register_operand" " f, f"))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " f, f, f, f"))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_scalar_narrow" @@ -7470,7 +7510,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,none,none")]) (define_expand "@pred_eqne<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -7516,24 +7557,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSF - (match_operand:<VEL> 5 "register_operand" " f, f")) - (match_operand:V_VLSF 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " f, f, f, f")) + (match_operand:V_VLSF 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_eqne<mode>_scalar_narrow" @@ -7554,7 +7596,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,none,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated floating-point merge @@ -7774,7 +7817,8 @@ [(set_attr "type" "vfncvtftoi") (set_attr "mode" "<VNCONVERT>") (set (attr "frm_mode") - (symbol_ref "riscv_vector::get_frm_mode (operands[8])"))]) + (symbol_ref "riscv_vector::get_frm_mode (operands[8])")) + (set_attr "group_overlap" "none,none,th,th,none,none")]) (define_insn "@pred_narrow_<fix_cvt><mode>" [(set (match_operand:<VNCONVERT> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -7816,7 +7860,8 @@ [(set_attr "type" "vfncvtitof") (set_attr "mode" "<VNCONVERT>") (set (attr "frm_mode") - (symbol_ref "riscv_vector::get_frm_mode (operands[8])"))]) + (symbol_ref "riscv_vector::get_frm_mode (operands[8])")) + (set_attr "group_overlap" "none,none,th,th,none,none")]) (define_insn "@pred_trunc<mode>" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -7839,7 +7884,8 @@ [(set_attr "type" "vfncvtftof") (set_attr "mode" "<V_DOUBLE_TRUNC>") (set (attr "frm_mode") - (symbol_ref "riscv_vector::get_frm_mode (operands[8])"))]) + (symbol_ref "riscv_vector::get_frm_mode (operands[8])")) + (set_attr "group_overlap" "none,none,th,th,none,none")]) (define_insn "@pred_rod_trunc<mode>" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: [PATCH v5] RISC-V: Fix register overlap issue for some xtheadvector instructions 2024-01-10 6:02 ` [PATCH v5] " Jun Sha (Joshua) @ 2024-01-10 6:37 ` juzhe.zhong 0 siblings, 0 replies; 133+ messages in thread From: juzhe.zhong @ 2024-01-10 6:37 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, cooper.joshua, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 38864 bytes --] + (and (eq_attr "group_overlap" "th") + (match_test "TARGET_XTHEADVECTOR")) + (const_string "no") + + (and (eq_attr "group_overlap" "rvv") + (match_test "TARGET_VECTOR && !TARGET_XTHEADVECTOR")) + (const_string "no") + ] Change it into: + (and (eq_attr "group_overlap" "thv_disabled") + (match_test "TARGET_XTHEADVECTOR")) + (const_string "no") + + (and (eq_attr "group_overlap" "rvv_disabled") + (match_test "TARGET_VECTOR && !TARGET_XTHEADVECTOR")) + (const_string "no") + ] juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-10 14:02 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Fix register overlap issue for some xtheadvector instructions For th.vmadc/th.vmsbc as well as narrowing arithmetic instructions and floating-point compare instructions, an illegal instruction exception will be raised if the destination vector register overlaps a source vector register group. To handle this issue, we use "group_overlap" and "enabled" attribute to disable some alternatives for xtheadvector. gcc/ChangeLog: * config/riscv/riscv.md (none,W21,W42,W84,W43,W86,W87,W0): (none,W21,W42,W84,W43,W86,W87,W0,th): Add group-overlap constraint for xtheadvector. * config/riscv/vector.md: Disable alternatives that destination register overlaps source register group for xtheadvector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config/riscv/riscv.md | 12 +- gcc/config/riscv/vector.md | 314 +++++++++++++++++++++---------------- 2 files changed, 190 insertions(+), 136 deletions(-) diff --git a/gcc/config/riscv/riscv.md b/gcc/config/riscv/riscv.md index 84212430dc0..2fe15fd7340 100644 --- a/gcc/config/riscv/riscv.md +++ b/gcc/config/riscv/riscv.md @@ -537,7 +537,7 @@ ;; Widening instructions have group-overlap constraints. Those are only ;; valid for certain register-group sizes. This attribute marks the ;; alternatives not matching the required register-group size as disabled. -(define_attr "group_overlap" "none,W21,W42,W84,W43,W86,W87,W0" +(define_attr "group_overlap" "none,W21,W42,W84,W43,W86,W87,W0,th,rvv" (const_string "none")) (define_attr "group_overlap_valid" "no,yes" @@ -576,7 +576,15 @@ (and (eq_attr "group_overlap" "W0") (match_test "riscv_get_v_regno_alignment (GET_MODE (operands[0])) > 1")) (const_string "no") - ] + + (and (eq_attr "group_overlap" "th") + (match_test "TARGET_XTHEADVECTOR")) + (const_string "no") + + (and (eq_attr "group_overlap" "rvv") + (match_test "TARGET_VECTOR && !TARGET_XTHEADVECTOR")) + (const_string "no") + ] (const_string "yes"))) ;; Attribute to control enable or disable instructions. diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index 3eb6daafbc2..cd83c1f3321 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -3260,7 +3260,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "th,none,none")]) (define_insn "@pred_msbc<mode>" [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr") @@ -3279,7 +3280,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "th,th,none")]) (define_insn "@pred_madc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3299,7 +3301,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "th,none")]) (define_insn "@pred_msbc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3319,7 +3322,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "th,none")]) (define_expand "@pred_madc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3368,7 +3372,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "th,none")]) (define_insn "*pred_madc<mode>_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3389,7 +3394,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "th,none")]) (define_expand "@pred_msbc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3438,7 +3444,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "th,none")]) (define_insn "*pred_msbc<mode>_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3459,7 +3466,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "th,none")]) (define_insn "@pred_madc<mode>_overflow" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr, &vr") @@ -3477,7 +3485,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "th,none,none")]) (define_insn "@pred_msbc<mode>_overflow" [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") @@ -3495,7 +3504,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "th,th,none,none")]) (define_insn "@pred_madc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3514,7 +3524,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "th,none")]) (define_insn "@pred_msbc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3533,7 +3544,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "th,none")]) (define_expand "@pred_madc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3580,7 +3592,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "th,none")]) (define_insn "*pred_madc<mode>_overflow_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3600,7 +3613,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "th,none")]) (define_expand "@pred_msbc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3647,7 +3661,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "th,none")]) (define_insn "*pred_msbc<mode>_overflow_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3667,7 +3682,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "th,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated integer unary operations @@ -3987,7 +4003,8 @@ "TARGET_VECTOR" "vn<insn>.w%o4\t%0,%3,%v4%p1" [(set_attr "type" "vnshift") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "group_overlap" "none,none,th,th,none,th,none,none,none,th,none,none")]) (define_insn "@pred_narrow_<optab><mode>_scalar" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -4008,7 +4025,8 @@ "TARGET_VECTOR" "vn<insn>.w%o4\t%0,%3,%4%p1" [(set_attr "type" "vnshift") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "group_overlap" "none,none,th,th,none,none")]) ;; vncvt.x.x.w (define_insn "@pred_trunc<mode>" @@ -4032,7 +4050,8 @@ (set_attr "vl_op_idx" "4") (set (attr "ta") (symbol_ref "riscv_vector::get_ta(operands[5])")) (set (attr "ma") (symbol_ref "riscv_vector::get_ma(operands[6])")) - (set (attr "avl_type_idx") (const_int 7))]) + (set (attr "avl_type_idx") (const_int 7)) + (set_attr "group_overlap" "none,none,th,th,none,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated fixed-point operations @@ -4438,7 +4457,8 @@ "TARGET_VECTOR" "vnclip<v_su>.w%o4\t%0,%3,%v4%p1" [(set_attr "type" "vnclip") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "group_overlap" "th,th,th,th,th,th,none,none,th,th,none,none")]) (define_insn "@pred_narrow_clip<v_su><mode>_scalar" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -4460,7 +4480,8 @@ "TARGET_VECTOR" "vnclip<v_su>.w%o4\t%0,%3,%4%p1" [(set_attr "type" "vnclip") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "group_overlap" "th,th,th,th,none,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated integer comparison operations @@ -4511,23 +4532,24 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr, &vr, &vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") - (match_operand 7 "const_int_operand" " i, i, i, i") - (match_operand 8 "const_int_operand" " i, i, i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK, rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i, i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_ltge_operator" - [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr") - (match_operand:V_VLSI 5 "vector_arith_operand" " vr, vr, vi, vi")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] + [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr, vr, vr, vr, vr") + (match_operand:V_VLSI 5 "vector_arith_operand" " vr, vr, vi, vi, vr, vr, vi, vi")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0, vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,th,th,rvv,rvv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_narrow" @@ -4547,7 +4569,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,th,th,th,th,none,none")]) (define_expand "@pred_ltge<mode>" [(set (match_operand:<VM> 0 "register_operand") @@ -4591,23 +4614,24 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_ltge<mode>" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr, &vr, &vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") - (match_operand 7 "const_int_operand" " i, i, i, i") - (match_operand 8 "const_int_operand" " i, i, i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK, rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i, i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "ltge_operator" - [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr") - (match_operand:V_VLSI 5 "vector_neg_arith_operand" " vr, vr, vj, vj")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] + [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr, vr, vr, vr, vr") + (match_operand:V_VLSI 5 "vector_neg_arith_operand" " vr, vr, vj, vj, vr, vr, vj, vj")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0, vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,th,th,rvv,rvv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_ltge<mode>_narrow" @@ -4627,7 +4651,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,th,th,th,th,none,none")]) (define_expand "@pred_cmp<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -4673,24 +4698,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_eqge_operator" - [(match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr") + [(match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSI_QHS - (match_operand:<VEL> 5 "register_operand" " r, r"))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r"))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_scalar_narrow" @@ -4711,7 +4737,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,none,none")]) (define_expand "@pred_eqne<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -4757,24 +4784,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSI_QHS - (match_operand:<VEL> 5 "register_operand" " r, r")) - (match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r")) + (match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_eqne<mode>_scalar_narrow" @@ -4795,7 +4823,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,none,none")]) ;; Handle GET_MODE_INNER (mode) = DImode. We need to split them since ;; we need to deal with SEW = 64 in RV32 system. @@ -4922,24 +4951,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_eqge_operator" - [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr") + [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSI_D - (match_operand:<VEL> 5 "register_operand" " r, r"))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r"))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_scalar_narrow" @@ -4960,28 +4990,30 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,none,none")]) ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSI_D - (match_operand:<VEL> 5 "register_operand" " r, r")) - (match_operand:V_VLSI_D 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r")) + (match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_eqne<mode>_scalar_narrow" @@ -5002,7 +5034,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,none,none")]) (define_insn "*pred_cmp<mode>_extended_scalar_merge_tie_mask" [(set (match_operand:<VM> 0 "register_operand" "=vm") @@ -5031,25 +5064,26 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_extended_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_eqge_operator" - [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr") + [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSI_D (sign_extend:<VEL> - (match_operand:<VSUBEL> 5 "register_operand" " r, r")))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VSUBEL> 5 "register_operand" " r, r, r, r")))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,rvv,rvv")]) (define_insn "*pred_cmp<mode>_extended_scalar_narrow" [(set (match_operand:<VM> 0 "register_operand" "=vm, vr, vr, &vr, &vr") @@ -5070,7 +5104,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,none,none")]) (define_insn "*pred_eqne<mode>_extended_scalar_merge_tie_mask" [(set (match_operand:<VM> 0 "register_operand" "=vm") @@ -5099,25 +5134,26 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_extended_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSI_D (sign_extend:<VEL> - (match_operand:<VSUBEL> 5 "register_operand" " r, r"))) - (match_operand:V_VLSI_D 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VSUBEL> 5 "register_operand" " r, r, r, r"))) + (match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,rvv,rvv")]) (define_insn "*pred_eqne<mode>_extended_scalar_narrow" [(set (match_operand:<VM> 0 "register_operand" "=vm, vr, vr, &vr, &vr") @@ -5138,7 +5174,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,none,none")]) ;; GE, vmsge.vx/vmsgeu.vx ;; @@ -7327,23 +7364,24 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "signed_order_operator" - [(match_operand:V_VLSF 4 "register_operand" " vr, vr") - (match_operand:V_VLSF 5 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + [(match_operand:V_VLSF 4 "register_operand" " vr, vr, vr, vr") + (match_operand:V_VLSF 5 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vmf%B3.vv\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,rvv,rvv")]) (define_insn "*pred_cmp<mode>_narrow_merge_tie_mask" [(set (match_operand:<VM> 0 "register_operand" "=vm") @@ -7386,7 +7424,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vmf%B3.vv\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,th,th,th,th,none,none")]) (define_expand "@pred_cmp<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -7432,24 +7471,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "signed_order_operator" - [(match_operand:V_VLSF 4 "register_operand" " vr, vr") + [(match_operand:V_VLSF 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSF - (match_operand:<VEL> 5 "register_operand" " f, f"))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " f, f, f, f"))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_scalar_narrow" @@ -7470,7 +7510,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,none,none")]) (define_expand "@pred_eqne<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -7516,24 +7557,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSF - (match_operand:<VEL> 5 "register_operand" " f, f")) - (match_operand:V_VLSF 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " f, f, f, f")) + (match_operand:V_VLSF 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "th,th,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_eqne<mode>_scalar_narrow" @@ -7554,7 +7596,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,th,th,none,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated floating-point merge @@ -7774,7 +7817,8 @@ [(set_attr "type" "vfncvtftoi") (set_attr "mode" "<VNCONVERT>") (set (attr "frm_mode") - (symbol_ref "riscv_vector::get_frm_mode (operands[8])"))]) + (symbol_ref "riscv_vector::get_frm_mode (operands[8])")) + (set_attr "group_overlap" "none,none,th,th,none,none")]) (define_insn "@pred_narrow_<fix_cvt><mode>" [(set (match_operand:<VNCONVERT> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -7816,7 +7860,8 @@ [(set_attr "type" "vfncvtitof") (set_attr "mode" "<VNCONVERT>") (set (attr "frm_mode") - (symbol_ref "riscv_vector::get_frm_mode (operands[8])"))]) + (symbol_ref "riscv_vector::get_frm_mode (operands[8])")) + (set_attr "group_overlap" "none,none,th,th,none,none")]) (define_insn "@pred_trunc<mode>" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -7839,7 +7884,8 @@ [(set_attr "type" "vfncvtftof") (set_attr "mode" "<V_DOUBLE_TRUNC>") (set (attr "frm_mode") - (symbol_ref "riscv_vector::get_frm_mode (operands[8])"))]) + (symbol_ref "riscv_vector::get_frm_mode (operands[8])")) + (set_attr "group_overlap" "none,none,th,th,none,none")]) (define_insn "@pred_rod_trunc<mode>" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* [PATCH v5] RISC-V: Fix register overlap issue for some xtheadvector instructions 2024-01-03 7:54 ` Jun Sha (Joshua) 2024-01-10 6:02 ` [PATCH v5] " Jun Sha (Joshua) @ 2024-01-10 6:51 ` Jun Sha (Joshua) 2024-01-10 6:53 ` juzhe.zhong ` (2 more replies) 1 sibling, 3 replies; 133+ messages in thread From: Jun Sha (Joshua) @ 2024-01-10 6:51 UTC (permalink / raw) To: gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, Jun Sha (Joshua), Jin Ma, Xianmiao Qu For th.vmadc/th.vmsbc as well as narrowing arithmetic instructions and floating-point compare instructions, an illegal instruction exception will be raised if the destination vector register overlaps a source vector register group. To handle this issue, we use "group_overlap" and "enabled" attribute to disable some alternatives for xtheadvector. gcc/ChangeLog: * config/riscv/riscv.md (none,W21,W42,W84,W43,W86,W87,W0): (none,W21,W42,W84,W43,W86,W87,W0,thv_disabled,rvv_disabled): Add group-overlap constraint for xtheadvector. * config/riscv/vector.md: Disable alternatives that destination register overlaps source register group for xtheadvector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config/riscv/riscv.md | 12 +- gcc/config/riscv/vector.md | 314 +++++++++++++++++++++---------------- 2 files changed, 190 insertions(+), 136 deletions(-) diff --git a/gcc/config/riscv/riscv.md b/gcc/config/riscv/riscv.md index 84212430dc0..411d1d17391 100644 --- a/gcc/config/riscv/riscv.md +++ b/gcc/config/riscv/riscv.md @@ -537,7 +537,7 @@ ;; Widening instructions have group-overlap constraints. Those are only ;; valid for certain register-group sizes. This attribute marks the ;; alternatives not matching the required register-group size as disabled. -(define_attr "group_overlap" "none,W21,W42,W84,W43,W86,W87,W0" +(define_attr "group_overlap" "none,W21,W42,W84,W43,W86,W87,W0,thv_disabled,rvv_disabled" (const_string "none")) (define_attr "group_overlap_valid" "no,yes" @@ -576,7 +576,15 @@ (and (eq_attr "group_overlap" "W0") (match_test "riscv_get_v_regno_alignment (GET_MODE (operands[0])) > 1")) (const_string "no") - ] + + (and (eq_attr "group_overlap" "thv_disabled") + (match_test "TARGET_XTHEADVECTOR")) + (const_string "no") + + (and (eq_attr "group_overlap" "rvv_disabled") + (match_test "TARGET_VECTOR && !TARGET_XTHEADVECTOR")) + (const_string "no") + ] (const_string "yes"))) ;; Attribute to control enable or disable instructions. diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index 3eb6daafbc2..4748ddd34a2 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -3260,7 +3260,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "thv_disabled,none,none")]) (define_insn "@pred_msbc<mode>" [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr") @@ -3279,7 +3280,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "thv_disabled,thv_disabled,none")]) (define_insn "@pred_madc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3299,7 +3301,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "thv_disabled,none")]) (define_insn "@pred_msbc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3319,7 +3322,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "thv_disabled,none")]) (define_expand "@pred_madc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3368,7 +3372,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "thv_disabled,none")]) (define_insn "*pred_madc<mode>_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3389,7 +3394,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "thv_disabled,none")]) (define_expand "@pred_msbc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3438,7 +3444,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "thv_disabled,none")]) (define_insn "*pred_msbc<mode>_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3459,7 +3466,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "thv_disabled,none")]) (define_insn "@pred_madc<mode>_overflow" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr, &vr") @@ -3477,7 +3485,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "thv_disabled,none,none")]) (define_insn "@pred_msbc<mode>_overflow" [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") @@ -3495,7 +3504,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "thv_disabled,thv_disabled,none,none")]) (define_insn "@pred_madc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3514,7 +3524,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "thv_disabled,none")]) (define_insn "@pred_msbc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3533,7 +3544,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "thv_disabled,none")]) (define_expand "@pred_madc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3580,7 +3592,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "thv_disabled,none")]) (define_insn "*pred_madc<mode>_overflow_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3600,7 +3613,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "thv_disabled,none")]) (define_expand "@pred_msbc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3647,7 +3661,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "thv_disabled,none")]) (define_insn "*pred_msbc<mode>_overflow_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3667,7 +3682,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "thv_disabled,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated integer unary operations @@ -3987,7 +4003,8 @@ "TARGET_VECTOR" "vn<insn>.w%o4\t%0,%3,%v4%p1" [(set_attr "type" "vnshift") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "group_overlap" "none,none,thv_disabled,thv_disabled,none,thv_disabled,none,none,none,thv_disabled,none,none")]) (define_insn "@pred_narrow_<optab><mode>_scalar" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -4008,7 +4025,8 @@ "TARGET_VECTOR" "vn<insn>.w%o4\t%0,%3,%4%p1" [(set_attr "type" "vnshift") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "group_overlap" "none,none,thv_disabled,thv_disabled,none,none")]) ;; vncvt.x.x.w (define_insn "@pred_trunc<mode>" @@ -4032,7 +4050,8 @@ (set_attr "vl_op_idx" "4") (set (attr "ta") (symbol_ref "riscv_vector::get_ta(operands[5])")) (set (attr "ma") (symbol_ref "riscv_vector::get_ma(operands[6])")) - (set (attr "avl_type_idx") (const_int 7))]) + (set (attr "avl_type_idx") (const_int 7)) + (set_attr "group_overlap" "none,none,thv_disabled,thv_disabled,none,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated fixed-point operations @@ -4438,7 +4457,8 @@ "TARGET_VECTOR" "vnclip<v_su>.w%o4\t%0,%3,%v4%p1" [(set_attr "type" "vnclip") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "group_overlap" "thv_disabled,thv_disabled,thv_disabled,thv_disabled,thv_disabled,thv_disabled,none,none,thv_disabled,thv_disabled,none,none")]) (define_insn "@pred_narrow_clip<v_su><mode>_scalar" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -4460,7 +4480,8 @@ "TARGET_VECTOR" "vnclip<v_su>.w%o4\t%0,%3,%4%p1" [(set_attr "type" "vnclip") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "group_overlap" "thv_disabled,thv_disabled,thv_disabled,thv_disabled,none,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated integer comparison operations @@ -4511,23 +4532,24 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr, &vr, &vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") - (match_operand 7 "const_int_operand" " i, i, i, i") - (match_operand 8 "const_int_operand" " i, i, i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK, rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i, i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_ltge_operator" - [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr") - (match_operand:V_VLSI 5 "vector_arith_operand" " vr, vr, vi, vi")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] + [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr, vr, vr, vr, vr") + (match_operand:V_VLSI 5 "vector_arith_operand" " vr, vr, vi, vi, vr, vr, vi, vi")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0, vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "thv_disabled,thv_disabled,thv_disabled,thv_disabled,rvv_disabled,rvv_disabled,rvv_disabled,rvv_disabled")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_narrow" @@ -4547,7 +4569,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,thv_disabled,thv_disabled,thv_disabled,thv_disabled,thv_disabled,thv_disabled,none,none")]) (define_expand "@pred_ltge<mode>" [(set (match_operand:<VM> 0 "register_operand") @@ -4591,23 +4614,24 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_ltge<mode>" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr, &vr, &vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") - (match_operand 7 "const_int_operand" " i, i, i, i") - (match_operand 8 "const_int_operand" " i, i, i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK, rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i, i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "ltge_operator" - [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr") - (match_operand:V_VLSI 5 "vector_neg_arith_operand" " vr, vr, vj, vj")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] + [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr, vr, vr, vr, vr") + (match_operand:V_VLSI 5 "vector_neg_arith_operand" " vr, vr, vj, vj, vr, vr, vj, vj")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0, vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "thv_disabled,thv_disabled,thv_disabled,thv_disabled,rvv_disabled,rvv_disabled,rvv_disabled,rvv_disabled")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_ltge<mode>_narrow" @@ -4627,7 +4651,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,thv_disabled,thv_disabled,thv_disabled,thv_disabled,thv_disabled,thv_disabled,none,none")]) (define_expand "@pred_cmp<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -4673,24 +4698,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_eqge_operator" - [(match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr") + [(match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSI_QHS - (match_operand:<VEL> 5 "register_operand" " r, r"))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r"))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "thv_disabled,thv_disabled,rvv_disabled,rvv_disabled")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_scalar_narrow" @@ -4711,7 +4737,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,thv_disabled,thv_disabled,none,none")]) (define_expand "@pred_eqne<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -4757,24 +4784,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSI_QHS - (match_operand:<VEL> 5 "register_operand" " r, r")) - (match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r")) + (match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "thv_disabled,thv_disabled,rvv_disabled,rvv_disabled")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_eqne<mode>_scalar_narrow" @@ -4795,7 +4823,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,thv_disabled,thv_disabled,none,none")]) ;; Handle GET_MODE_INNER (mode) = DImode. We need to split them since ;; we need to deal with SEW = 64 in RV32 system. @@ -4922,24 +4951,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_eqge_operator" - [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr") + [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSI_D - (match_operand:<VEL> 5 "register_operand" " r, r"))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r"))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "thv_disabled,thv_disabled,rvv_disabled,rvv_disabled")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_scalar_narrow" @@ -4960,28 +4990,30 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,thv_disabled,thv_disabled,none,none")]) ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSI_D - (match_operand:<VEL> 5 "register_operand" " r, r")) - (match_operand:V_VLSI_D 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r")) + (match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "thv_disabled,thv_disabled,rvv_disabled,rvv_disabled")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_eqne<mode>_scalar_narrow" @@ -5002,7 +5034,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,thv_disabled,thv_disabled,none,none")]) (define_insn "*pred_cmp<mode>_extended_scalar_merge_tie_mask" [(set (match_operand:<VM> 0 "register_operand" "=vm") @@ -5031,25 +5064,26 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_extended_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_eqge_operator" - [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr") + [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSI_D (sign_extend:<VEL> - (match_operand:<VSUBEL> 5 "register_operand" " r, r")))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VSUBEL> 5 "register_operand" " r, r, r, r")))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "thv_disabled,thv_disabled,rvv_disabled,rvv_disabled")]) (define_insn "*pred_cmp<mode>_extended_scalar_narrow" [(set (match_operand:<VM> 0 "register_operand" "=vm, vr, vr, &vr, &vr") @@ -5070,7 +5104,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,thv_disabled,thv_disabled,none,none")]) (define_insn "*pred_eqne<mode>_extended_scalar_merge_tie_mask" [(set (match_operand:<VM> 0 "register_operand" "=vm") @@ -5099,25 +5134,26 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_extended_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSI_D (sign_extend:<VEL> - (match_operand:<VSUBEL> 5 "register_operand" " r, r"))) - (match_operand:V_VLSI_D 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VSUBEL> 5 "register_operand" " r, r, r, r"))) + (match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "thv_disabled,thv_disabled,rvv_disabled,rvv_disabled")]) (define_insn "*pred_eqne<mode>_extended_scalar_narrow" [(set (match_operand:<VM> 0 "register_operand" "=vm, vr, vr, &vr, &vr") @@ -5138,7 +5174,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,thv_disabled,thv_disabled,none,none")]) ;; GE, vmsge.vx/vmsgeu.vx ;; @@ -7327,23 +7364,24 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "signed_order_operator" - [(match_operand:V_VLSF 4 "register_operand" " vr, vr") - (match_operand:V_VLSF 5 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + [(match_operand:V_VLSF 4 "register_operand" " vr, vr, vr, vr") + (match_operand:V_VLSF 5 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vmf%B3.vv\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "thv_disabled,thv_disabled,rvv_disabled,rvv_disabled")]) (define_insn "*pred_cmp<mode>_narrow_merge_tie_mask" [(set (match_operand:<VM> 0 "register_operand" "=vm") @@ -7386,7 +7424,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vmf%B3.vv\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,thv_disabled,thv_disabled,thv_disabled,thv_disabled,thv_disabled,thv_disabled,none,none")]) (define_expand "@pred_cmp<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -7432,24 +7471,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "signed_order_operator" - [(match_operand:V_VLSF 4 "register_operand" " vr, vr") + [(match_operand:V_VLSF 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSF - (match_operand:<VEL> 5 "register_operand" " f, f"))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " f, f, f, f"))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "thv_disabled,thv_disabled,rvv_disabled,rvv_disabled")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_scalar_narrow" @@ -7470,7 +7510,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,thv_disabled,thv_disabled,none,none")]) (define_expand "@pred_eqne<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -7516,24 +7557,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSF - (match_operand:<VEL> 5 "register_operand" " f, f")) - (match_operand:V_VLSF 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " f, f, f, f")) + (match_operand:V_VLSF 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "thv_disabled,thv_disabled,rvv_disabled,rvv_disabled")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_eqne<mode>_scalar_narrow" @@ -7554,7 +7596,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,thv_disabled,thv_disabled,none,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated floating-point merge @@ -7774,7 +7817,8 @@ [(set_attr "type" "vfncvtftoi") (set_attr "mode" "<VNCONVERT>") (set (attr "frm_mode") - (symbol_ref "riscv_vector::get_frm_mode (operands[8])"))]) + (symbol_ref "riscv_vector::get_frm_mode (operands[8])")) + (set_attr "group_overlap" "none,none,thv_disabled,thv_disabled,none,none")]) (define_insn "@pred_narrow_<fix_cvt><mode>" [(set (match_operand:<VNCONVERT> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -7816,7 +7860,8 @@ [(set_attr "type" "vfncvtitof") (set_attr "mode" "<VNCONVERT>") (set (attr "frm_mode") - (symbol_ref "riscv_vector::get_frm_mode (operands[8])"))]) + (symbol_ref "riscv_vector::get_frm_mode (operands[8])")) + (set_attr "group_overlap" "none,none,thv_disabled,thv_disabled,none,none")]) (define_insn "@pred_trunc<mode>" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -7839,7 +7884,8 @@ [(set_attr "type" "vfncvtftof") (set_attr "mode" "<V_DOUBLE_TRUNC>") (set (attr "frm_mode") - (symbol_ref "riscv_vector::get_frm_mode (operands[8])"))]) + (symbol_ref "riscv_vector::get_frm_mode (operands[8])")) + (set_attr "group_overlap" "none,none,thv_disabled,thv_disabled,none,none")]) (define_insn "@pred_rod_trunc<mode>" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: [PATCH v5] RISC-V: Fix register overlap issue for some xtheadvector instructions 2024-01-10 6:51 ` Jun Sha (Joshua) @ 2024-01-10 6:53 ` juzhe.zhong 2024-01-10 13:36 ` Robin Dapp 2024-01-11 2:39 ` Jun Sha (Joshua) 2 siblings, 0 replies; 133+ messages in thread From: juzhe.zhong @ 2024-01-10 6:53 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, cooper.joshua, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 39751 bytes --] LGTM from myside. Give another a few more days that some one want to chime in. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-10 14:51 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Fix register overlap issue for some xtheadvector instructions For th.vmadc/th.vmsbc as well as narrowing arithmetic instructions and floating-point compare instructions, an illegal instruction exception will be raised if the destination vector register overlaps a source vector register group. To handle this issue, we use "group_overlap" and "enabled" attribute to disable some alternatives for xtheadvector. gcc/ChangeLog: * config/riscv/riscv.md (none,W21,W42,W84,W43,W86,W87,W0): (none,W21,W42,W84,W43,W86,W87,W0,thv_disabled,rvv_disabled): Add group-overlap constraint for xtheadvector. * config/riscv/vector.md: Disable alternatives that destination register overlaps source register group for xtheadvector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config/riscv/riscv.md | 12 +- gcc/config/riscv/vector.md | 314 +++++++++++++++++++++---------------- 2 files changed, 190 insertions(+), 136 deletions(-) diff --git a/gcc/config/riscv/riscv.md b/gcc/config/riscv/riscv.md index 84212430dc0..411d1d17391 100644 --- a/gcc/config/riscv/riscv.md +++ b/gcc/config/riscv/riscv.md @@ -537,7 +537,7 @@ ;; Widening instructions have group-overlap constraints. Those are only ;; valid for certain register-group sizes. This attribute marks the ;; alternatives not matching the required register-group size as disabled. -(define_attr "group_overlap" "none,W21,W42,W84,W43,W86,W87,W0" +(define_attr "group_overlap" "none,W21,W42,W84,W43,W86,W87,W0,thv_disabled,rvv_disabled" (const_string "none")) (define_attr "group_overlap_valid" "no,yes" @@ -576,7 +576,15 @@ (and (eq_attr "group_overlap" "W0") (match_test "riscv_get_v_regno_alignment (GET_MODE (operands[0])) > 1")) (const_string "no") - ] + + (and (eq_attr "group_overlap" "thv_disabled") + (match_test "TARGET_XTHEADVECTOR")) + (const_string "no") + + (and (eq_attr "group_overlap" "rvv_disabled") + (match_test "TARGET_VECTOR && !TARGET_XTHEADVECTOR")) + (const_string "no") + ] (const_string "yes"))) ;; Attribute to control enable or disable instructions. diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index 3eb6daafbc2..4748ddd34a2 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -3260,7 +3260,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "thv_disabled,none,none")]) (define_insn "@pred_msbc<mode>" [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr") @@ -3279,7 +3280,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "thv_disabled,thv_disabled,none")]) (define_insn "@pred_madc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3299,7 +3301,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "thv_disabled,none")]) (define_insn "@pred_msbc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3319,7 +3322,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "thv_disabled,none")]) (define_expand "@pred_madc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3368,7 +3372,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "thv_disabled,none")]) (define_insn "*pred_madc<mode>_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3389,7 +3394,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "thv_disabled,none")]) (define_expand "@pred_msbc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3438,7 +3444,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "thv_disabled,none")]) (define_insn "*pred_msbc<mode>_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3459,7 +3466,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "group_overlap" "thv_disabled,none")]) (define_insn "@pred_madc<mode>_overflow" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr, &vr") @@ -3477,7 +3485,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "thv_disabled,none,none")]) (define_insn "@pred_msbc<mode>_overflow" [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") @@ -3495,7 +3504,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "thv_disabled,thv_disabled,none,none")]) (define_insn "@pred_madc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3514,7 +3524,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "thv_disabled,none")]) (define_insn "@pred_msbc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3533,7 +3544,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "thv_disabled,none")]) (define_expand "@pred_madc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3580,7 +3592,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "thv_disabled,none")]) (define_insn "*pred_madc<mode>_overflow_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3600,7 +3613,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "thv_disabled,none")]) (define_expand "@pred_msbc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3647,7 +3661,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "thv_disabled,none")]) (define_insn "*pred_msbc<mode>_overflow_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3667,7 +3682,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "group_overlap" "thv_disabled,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated integer unary operations @@ -3987,7 +4003,8 @@ "TARGET_VECTOR" "vn<insn>.w%o4\t%0,%3,%v4%p1" [(set_attr "type" "vnshift") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "group_overlap" "none,none,thv_disabled,thv_disabled,none,thv_disabled,none,none,none,thv_disabled,none,none")]) (define_insn "@pred_narrow_<optab><mode>_scalar" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -4008,7 +4025,8 @@ "TARGET_VECTOR" "vn<insn>.w%o4\t%0,%3,%4%p1" [(set_attr "type" "vnshift") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "group_overlap" "none,none,thv_disabled,thv_disabled,none,none")]) ;; vncvt.x.x.w (define_insn "@pred_trunc<mode>" @@ -4032,7 +4050,8 @@ (set_attr "vl_op_idx" "4") (set (attr "ta") (symbol_ref "riscv_vector::get_ta(operands[5])")) (set (attr "ma") (symbol_ref "riscv_vector::get_ma(operands[6])")) - (set (attr "avl_type_idx") (const_int 7))]) + (set (attr "avl_type_idx") (const_int 7)) + (set_attr "group_overlap" "none,none,thv_disabled,thv_disabled,none,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated fixed-point operations @@ -4438,7 +4457,8 @@ "TARGET_VECTOR" "vnclip<v_su>.w%o4\t%0,%3,%v4%p1" [(set_attr "type" "vnclip") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "group_overlap" "thv_disabled,thv_disabled,thv_disabled,thv_disabled,thv_disabled,thv_disabled,none,none,thv_disabled,thv_disabled,none,none")]) (define_insn "@pred_narrow_clip<v_su><mode>_scalar" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -4460,7 +4480,8 @@ "TARGET_VECTOR" "vnclip<v_su>.w%o4\t%0,%3,%4%p1" [(set_attr "type" "vnclip") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "group_overlap" "thv_disabled,thv_disabled,thv_disabled,thv_disabled,none,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated integer comparison operations @@ -4511,23 +4532,24 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr, &vr, &vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") - (match_operand 7 "const_int_operand" " i, i, i, i") - (match_operand 8 "const_int_operand" " i, i, i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK, rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i, i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_ltge_operator" - [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr") - (match_operand:V_VLSI 5 "vector_arith_operand" " vr, vr, vi, vi")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] + [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr, vr, vr, vr, vr") + (match_operand:V_VLSI 5 "vector_arith_operand" " vr, vr, vi, vi, vr, vr, vi, vi")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0, vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "thv_disabled,thv_disabled,thv_disabled,thv_disabled,rvv_disabled,rvv_disabled,rvv_disabled,rvv_disabled")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_narrow" @@ -4547,7 +4569,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,thv_disabled,thv_disabled,thv_disabled,thv_disabled,thv_disabled,thv_disabled,none,none")]) (define_expand "@pred_ltge<mode>" [(set (match_operand:<VM> 0 "register_operand") @@ -4591,23 +4614,24 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_ltge<mode>" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr, &vr, &vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") - (match_operand 7 "const_int_operand" " i, i, i, i") - (match_operand 8 "const_int_operand" " i, i, i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK, rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i, i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "ltge_operator" - [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr") - (match_operand:V_VLSI 5 "vector_neg_arith_operand" " vr, vr, vj, vj")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] + [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr, vr, vr, vr, vr") + (match_operand:V_VLSI 5 "vector_neg_arith_operand" " vr, vr, vj, vj, vr, vr, vj, vj")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0, vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "thv_disabled,thv_disabled,thv_disabled,thv_disabled,rvv_disabled,rvv_disabled,rvv_disabled,rvv_disabled")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_ltge<mode>_narrow" @@ -4627,7 +4651,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,thv_disabled,thv_disabled,thv_disabled,thv_disabled,thv_disabled,thv_disabled,none,none")]) (define_expand "@pred_cmp<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -4673,24 +4698,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_eqge_operator" - [(match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr") + [(match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSI_QHS - (match_operand:<VEL> 5 "register_operand" " r, r"))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r"))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "thv_disabled,thv_disabled,rvv_disabled,rvv_disabled")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_scalar_narrow" @@ -4711,7 +4737,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,thv_disabled,thv_disabled,none,none")]) (define_expand "@pred_eqne<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -4757,24 +4784,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSI_QHS - (match_operand:<VEL> 5 "register_operand" " r, r")) - (match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r")) + (match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "thv_disabled,thv_disabled,rvv_disabled,rvv_disabled")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_eqne<mode>_scalar_narrow" @@ -4795,7 +4823,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,thv_disabled,thv_disabled,none,none")]) ;; Handle GET_MODE_INNER (mode) = DImode. We need to split them since ;; we need to deal with SEW = 64 in RV32 system. @@ -4922,24 +4951,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_eqge_operator" - [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr") + [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSI_D - (match_operand:<VEL> 5 "register_operand" " r, r"))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r"))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "thv_disabled,thv_disabled,rvv_disabled,rvv_disabled")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_scalar_narrow" @@ -4960,28 +4990,30 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,thv_disabled,thv_disabled,none,none")]) ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSI_D - (match_operand:<VEL> 5 "register_operand" " r, r")) - (match_operand:V_VLSI_D 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r")) + (match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "thv_disabled,thv_disabled,rvv_disabled,rvv_disabled")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_eqne<mode>_scalar_narrow" @@ -5002,7 +5034,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,thv_disabled,thv_disabled,none,none")]) (define_insn "*pred_cmp<mode>_extended_scalar_merge_tie_mask" [(set (match_operand:<VM> 0 "register_operand" "=vm") @@ -5031,25 +5064,26 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_extended_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_eqge_operator" - [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr") + [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSI_D (sign_extend:<VEL> - (match_operand:<VSUBEL> 5 "register_operand" " r, r")))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VSUBEL> 5 "register_operand" " r, r, r, r")))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "thv_disabled,thv_disabled,rvv_disabled,rvv_disabled")]) (define_insn "*pred_cmp<mode>_extended_scalar_narrow" [(set (match_operand:<VM> 0 "register_operand" "=vm, vr, vr, &vr, &vr") @@ -5070,7 +5104,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,thv_disabled,thv_disabled,none,none")]) (define_insn "*pred_eqne<mode>_extended_scalar_merge_tie_mask" [(set (match_operand:<VM> 0 "register_operand" "=vm") @@ -5099,25 +5134,26 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_extended_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSI_D (sign_extend:<VEL> - (match_operand:<VSUBEL> 5 "register_operand" " r, r"))) - (match_operand:V_VLSI_D 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VSUBEL> 5 "register_operand" " r, r, r, r"))) + (match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "thv_disabled,thv_disabled,rvv_disabled,rvv_disabled")]) (define_insn "*pred_eqne<mode>_extended_scalar_narrow" [(set (match_operand:<VM> 0 "register_operand" "=vm, vr, vr, &vr, &vr") @@ -5138,7 +5174,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,thv_disabled,thv_disabled,none,none")]) ;; GE, vmsge.vx/vmsgeu.vx ;; @@ -7327,23 +7364,24 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "signed_order_operator" - [(match_operand:V_VLSF 4 "register_operand" " vr, vr") - (match_operand:V_VLSF 5 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + [(match_operand:V_VLSF 4 "register_operand" " vr, vr, vr, vr") + (match_operand:V_VLSF 5 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vmf%B3.vv\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "thv_disabled,thv_disabled,rvv_disabled,rvv_disabled")]) (define_insn "*pred_cmp<mode>_narrow_merge_tie_mask" [(set (match_operand:<VM> 0 "register_operand" "=vm") @@ -7386,7 +7424,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vmf%B3.vv\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,thv_disabled,thv_disabled,thv_disabled,thv_disabled,thv_disabled,thv_disabled,none,none")]) (define_expand "@pred_cmp<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -7432,24 +7471,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "signed_order_operator" - [(match_operand:V_VLSF 4 "register_operand" " vr, vr") + [(match_operand:V_VLSF 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSF - (match_operand:<VEL> 5 "register_operand" " f, f"))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " f, f, f, f"))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "thv_disabled,thv_disabled,rvv_disabled,rvv_disabled")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_scalar_narrow" @@ -7470,7 +7510,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,thv_disabled,thv_disabled,none,none")]) (define_expand "@pred_eqne<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -7516,24 +7557,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSF - (match_operand:<VEL> 5 "register_operand" " f, f")) - (match_operand:V_VLSF 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " f, f, f, f")) + (match_operand:V_VLSF 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "thv_disabled,thv_disabled,rvv_disabled,rvv_disabled")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_eqne<mode>_scalar_narrow" @@ -7554,7 +7596,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "group_overlap" "none,thv_disabled,thv_disabled,none,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated floating-point merge @@ -7774,7 +7817,8 @@ [(set_attr "type" "vfncvtftoi") (set_attr "mode" "<VNCONVERT>") (set (attr "frm_mode") - (symbol_ref "riscv_vector::get_frm_mode (operands[8])"))]) + (symbol_ref "riscv_vector::get_frm_mode (operands[8])")) + (set_attr "group_overlap" "none,none,thv_disabled,thv_disabled,none,none")]) (define_insn "@pred_narrow_<fix_cvt><mode>" [(set (match_operand:<VNCONVERT> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -7816,7 +7860,8 @@ [(set_attr "type" "vfncvtitof") (set_attr "mode" "<VNCONVERT>") (set (attr "frm_mode") - (symbol_ref "riscv_vector::get_frm_mode (operands[8])"))]) + (symbol_ref "riscv_vector::get_frm_mode (operands[8])")) + (set_attr "group_overlap" "none,none,thv_disabled,thv_disabled,none,none")]) (define_insn "@pred_trunc<mode>" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -7839,7 +7884,8 @@ [(set_attr "type" "vfncvtftof") (set_attr "mode" "<V_DOUBLE_TRUNC>") (set (attr "frm_mode") - (symbol_ref "riscv_vector::get_frm_mode (operands[8])"))]) + (symbol_ref "riscv_vector::get_frm_mode (operands[8])")) + (set_attr "group_overlap" "none,none,thv_disabled,thv_disabled,none,none")]) (define_insn "@pred_rod_trunc<mode>" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: [PATCH v5] RISC-V: Fix register overlap issue for some xtheadvector instructions 2024-01-10 6:51 ` Jun Sha (Joshua) 2024-01-10 6:53 ` juzhe.zhong @ 2024-01-10 13:36 ` Robin Dapp 2024-01-10 13:43 ` 钟居哲 2024-01-11 2:39 ` Jun Sha (Joshua) 2 siblings, 1 reply; 133+ messages in thread From: Robin Dapp @ 2024-01-10 13:36 UTC (permalink / raw) To: Jun Sha (Joshua), gcc-patches Cc: rdapp.gcc, jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, Jin Ma, Xianmiao Qu Hi Joshua, > For th.vmadc/th.vmsbc as well as narrowing arithmetic instructions > and floating-point compare instructions, an illegal instruction > exception will be raised if the destination vector register overlaps > a source vector register group. > > To handle this issue, we use "group_overlap" and "enabled" attribute > to disable some alternatives for xtheadvector. > ;; Widening instructions have group-overlap constraints. Those are only > ;; valid for certain register-group sizes. This attribute marks the > ;; alternatives not matching the required register-group size as disabled. > -(define_attr "group_overlap" "none,W21,W42,W84,W43,W86,W87,W0" > +(define_attr "group_overlap" "none,W21,W42,W84,W43,W86,W87,W0,thv_disabled,rvv_disabled" > (const_string "none")) I realize there have been some discussions before but I find the naming misleading. The group_overlap attribute is supposed to specify whether groups overlap (and mark the respective alternatives accepting only this overlap). Then we check if the groups overlap and disable all non-matching alternatives. "none" i.e. "no overlap" always matches. Your first goal seems to be to disable existing non-early-clobber alternatives for thv. For this, maybe "full", "same" (or "any"?) would work? Please also add a comment in group_overlap_valid then that we need not actually check for register equality. For the other insns, I wonder if we could get away with not really disabling the newly added early-clobber alternatives for RVV but just disparaging ("?") them? That way we could re-use "full" for the thv-disabled alternatives and "none" for the newly added ones. ("none" will still be misleading then, though :/) If this doesn't work or others feel the separation is not strict enough, I'd prefer a separate attribute rather than overloading group_overlap. Maybe something like "spec_restriction" or similar with two values "rvv" and "thv"? Regards Robin ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: Re: [PATCH v5] RISC-V: Fix register overlap issue for some xtheadvector instructions 2024-01-10 13:36 ` Robin Dapp @ 2024-01-10 13:43 ` 钟居哲 2024-01-11 2:40 ` Re：Re: " joshua 0 siblings, 1 reply; 133+ messages in thread From: 钟居哲 @ 2024-01-10 13:43 UTC (permalink / raw) To: rdapp.gcc, cooper.joshua, gcc-patches Cc: rdapp.gcc, jim.wilson.gcc, palmer, andrew, philipp.tomsich, Jeff Law, Christoph Müllner, jinma, Cooper Qu [-- Attachment #1: Type: text/plain, Size: 3224 bytes --] >> For the other insns, I wonder if we could get away with not really >>disabling the newly added early-clobber alternatives for RVV but >>just disparaging ("?") them? That way we could re-use "full" for >>the thv-disabled alternatives and "none" for the newly added ones. >>("none" will still be misleading then, though :/) I prefer to disable those early-clobber alternatives added of theadvector for RVV, since disparage still make RA possible reaches the early clobber alternatives. >>If this doesn't work or others feel the separation is not strict >>enough, I'd prefer a separate attribute rather than overloading >>group_overlap. Maybe something like "spec_restriction" or similar >>with two values "rvv" and "thv"? I like this idea, it makes more sense to me. So I think it's better to add an attribute to disable alternative for theadvector or RVV1.0. juzhe.zhong@rivai.ai From: Robin Dapp Date: 2024-01-10 21:36 To: Jun Sha (Joshua); gcc-patches CC: rdapp.gcc; jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jin Ma; Xianmiao Qu Subject: Re: [PATCH v5] RISC-V: Fix register overlap issue for some xtheadvector instructions Hi Joshua, > For th.vmadc/th.vmsbc as well as narrowing arithmetic instructions > and floating-point compare instructions, an illegal instruction > exception will be raised if the destination vector register overlaps > a source vector register group. > > To handle this issue, we use "group_overlap" and "enabled" attribute > to disable some alternatives for xtheadvector. > ;; Widening instructions have group-overlap constraints. Those are only > ;; valid for certain register-group sizes. This attribute marks the > ;; alternatives not matching the required register-group size as disabled. > -(define_attr "group_overlap" "none,W21,W42,W84,W43,W86,W87,W0" > +(define_attr "group_overlap" "none,W21,W42,W84,W43,W86,W87,W0,thv_disabled,rvv_disabled" > (const_string "none")) I realize there have been some discussions before but I find the naming misleading. The group_overlap attribute is supposed to specify whether groups overlap (and mark the respective alternatives accepting only this overlap). Then we check if the groups overlap and disable all non-matching alternatives. "none" i.e. "no overlap" always matches. Your first goal seems to be to disable existing non-early-clobber alternatives for thv. For this, maybe "full", "same" (or "any"?) would work? Please also add a comment in group_overlap_valid then that we need not actually check for register equality. For the other insns, I wonder if we could get away with not really disabling the newly added early-clobber alternatives for RVV but just disparaging ("?") them? That way we could re-use "full" for the thv-disabled alternatives and "none" for the newly added ones. ("none" will still be misleading then, though :/) If this doesn't work or others feel the separation is not strict enough, I'd prefer a separate attribute rather than overloading group_overlap. Maybe something like "spec_restriction" or similar with two values "rvv" and "thv"? Regards Robin ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re：Re: [PATCH v5] RISC-V: Fix register overlap issue for some xtheadvector instructions 2024-01-10 13:43 ` 钟居哲 @ 2024-01-11 2:40 ` joshua 0 siblings, 0 replies; 133+ messages in thread From: joshua @ 2024-01-11 2:40 UTC (permalink / raw) To: 钟居哲, rdapp.gcc, gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, Jeff Law, Christoph Müllner, jinma, Cooper Qu Hi Robin, Thank you for your suggestions! The patch has been updated by adding a new attribute to disable alternative for xtheadvector or RVV1.0 instead of overlaoding group_overlap. Joshua ------------------------------------------------------------------ 发件人：钟居哲 <juzhe.zhong@rivai.ai> 发送时间：2024年1月10日(星期三) 21:43 收件人："rdapp.gcc"<rdapp.gcc@gmail.com>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送："rdapp.gcc"<rdapp.gcc@gmail.com>; "jim.wilson.gcc"<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; Jeff Law<jeffreyalaw@gmail.com>; "Christoph Müllner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; Cooper Qu<cooper.qu@linux.alibaba.com> 主 题：Re: Re: [PATCH v5] RISC-V: Fix register overlap issue for some xtheadvector instructions >> For the other insns, I wonder if we could get away with not really >>disabling the newly added early-clobber alternatives for RVV but >>just disparaging ("?") them? That way we could re-use "full" for >>the thv-disabled alternatives and "none" for the newly added ones. >>("none" will still be misleading then, though :/) I prefer to disable those early-clobber alternatives added of theadvector for RVV, since disparage still make RA possible reaches the early clobber alternatives. >>If this doesn't work or others feel the separation is not strict >>enough, I'd prefer a separate attribute rather than overloading >>group_overlap. Maybe something like "spec_restriction" or similar >>with two values "rvv" and "thv"? I like this idea, it makes more sense to me. So I think it's better to add an attribute to disable alternative for theadvector or RVV1.0. juzhe.zhong@rivai.ai From: Robin Dapp Date: 2024-01-10 21:36 To: Jun Sha (Joshua); gcc-patches CC: rdapp.gcc; jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jin Ma; Xianmiao Qu Subject: Re: [PATCH v5] RISC-V: Fix register overlap issue for some xtheadvector instructions Hi Joshua, > For th.vmadc/th.vmsbc as well as narrowing arithmetic instructions > and floating-point compare instructions, an illegal instruction > exception will be raised if the destination vector register overlaps > a source vector register group. > > To handle this issue, we use "group_overlap" and "enabled" attribute > to disable some alternatives for xtheadvector. > ;; Widening instructions have group-overlap constraints. Those are only > ;; valid for certain register-group sizes. This attribute marks the > ;; alternatives not matching the required register-group size as disabled. > -(define_attr "group_overlap" "none,W21,W42,W84,W43,W86,W87,W0" > +(define_attr "group_overlap" "none,W21,W42,W84,W43,W86,W87,W0,thv_disabled,rvv_disabled" > (const_string "none")) I realize there have been some discussions before but I find the naming misleading. The group_overlap attribute is supposed to specify whether groups overlap (and mark the respective alternatives accepting only this overlap). Then we check if the groups overlap and disable all non-matching alternatives. "none" i.e. "no overlap" always matches. Your first goal seems to be to disable existing non-early-clobber alternatives for thv. For this, maybe "full", "same" (or "any"?) would work? Please also add a comment in group_overlap_valid then that we need not actually check for register equality. For the other insns, I wonder if we could get away with not really disabling the newly added early-clobber alternatives for RVV but just disparaging ("?") them? That way we could re-use "full" for the thv-disabled alternatives and "none" for the newly added ones. ("none" will still be misleading then, though :/) If this doesn't work or others feel the separation is not strict enough, I'd prefer a separate attribute rather than overloading group_overlap. Maybe something like "spec_restriction" or similar with two values "rvv" and "thv"? Regards Robin ^ permalink raw reply [flat|nested] 133+ messages in thread
* [PATCH v5] RISC-V: Fix register overlap issue for some xtheadvector instructions 2024-01-10 6:51 ` Jun Sha (Joshua) 2024-01-10 6:53 ` juzhe.zhong 2024-01-10 13:36 ` Robin Dapp @ 2024-01-11 2:39 ` Jun Sha (Joshua) 2024-01-11 2:46 ` juzhe.zhong 2 siblings, 1 reply; 133+ messages in thread From: Jun Sha (Joshua) @ 2024-01-11 2:39 UTC (permalink / raw) To: gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, Jun Sha (Joshua), Jin Ma, Xianmiao Qu For th.vmadc/th.vmsbc as well as narrowing arithmetic instructions and floating-point compare instructions, an illegal instruction exception will be raised if the destination vector register overlaps a source vector register group. To handle this issue, we add an attribute "spec_restriction" to disable some alternatives for xtheadvector. gcc/ChangeLog: * config/riscv/riscv.md (none,thv,rvv): (no,yes): Add an attribute to disable alternative for xtheadvector or RVV1.0. * config/riscv/vector.md: Disable alternatives that destination register overlaps source register group for xtheadvector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config/riscv/riscv.md | 22 +++ gcc/config/riscv/vector.md | 314 +++++++++++++++++++++---------------- 2 files changed, 202 insertions(+), 134 deletions(-) diff --git a/gcc/config/riscv/riscv.md b/gcc/config/riscv/riscv.md index 84212430dc0..23fc32d5cb2 100644 --- a/gcc/config/riscv/riscv.md +++ b/gcc/config/riscv/riscv.md @@ -579,6 +579,25 @@ ] (const_string "yes"))) +;; This attribute marks the alternatives not matching the constraints +;; described in spec as disabled. +(define_attr "spec_restriction" "none,thv,rvv" + (const_string "none")) + +(define_attr "spec_restriction_disabled" "no,yes" + (cond [(eq_attr "spec_restriction" "none") + (const_string "no") + + (and (eq_attr "spec_restriction" "thv") + (match_test "TARGET_XTHEADVECTOR")) + (const_string "yes") + + (and (eq_attr "spec_restriction" "rvv") + (match_test "TARGET_VECTOR && !TARGET_XTHEADVECTOR")) + (const_string "yes") + ] + (const_string "no"))) + ;; Attribute to control enable or disable instructions. (define_attr "enabled" "no,yes" (cond [ @@ -590,6 +609,9 @@ (eq_attr "group_overlap_valid" "no") (const_string "no") + + (eq_attr "spec_restriction_disabled" "yes") + (const_string "no") ] (const_string "yes"))) diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index 3eb6daafbc2..c79416cf0d3 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -3260,7 +3260,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "spec_restriction" "thv,none,none")]) (define_insn "@pred_msbc<mode>" [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr") @@ -3279,7 +3280,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "spec_restriction" "thv,thv,none")]) (define_insn "@pred_madc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3299,7 +3301,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "spec_restriction" "thv,none")]) (define_insn "@pred_msbc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3319,7 +3322,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "spec_restriction" "thv,none")]) (define_expand "@pred_madc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3368,7 +3372,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "spec_restriction" "thv,none")]) (define_insn "*pred_madc<mode>_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3389,7 +3394,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "spec_restriction" "thv,none")]) (define_expand "@pred_msbc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3438,7 +3444,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "spec_restriction" "thv,none")]) (define_insn "*pred_msbc<mode>_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3459,7 +3466,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "spec_restriction" "thv,none")]) (define_insn "@pred_madc<mode>_overflow" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr, &vr") @@ -3477,7 +3485,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "spec_restriction" "thv,none,none")]) (define_insn "@pred_msbc<mode>_overflow" [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") @@ -3495,7 +3504,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "spec_restriction" "thv,thv,none,none")]) (define_insn "@pred_madc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3514,7 +3524,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "spec_restriction" "thv,none")]) (define_insn "@pred_msbc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3533,7 +3544,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "spec_restriction" "thv,none")]) (define_expand "@pred_madc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3580,7 +3592,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "spec_restriction" "thv,none")]) (define_insn "*pred_madc<mode>_overflow_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3600,7 +3613,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "spec_restriction" "thv,none")]) (define_expand "@pred_msbc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3647,7 +3661,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "spec_restriction" "thv,none")]) (define_insn "*pred_msbc<mode>_overflow_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3667,7 +3682,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "spec_restriction" "thv,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated integer unary operations @@ -3987,7 +4003,8 @@ "TARGET_VECTOR" "vn<insn>.w%o4\t%0,%3,%v4%p1" [(set_attr "type" "vnshift") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "spec_restriction" "none,none,thv,thv,none,thv,none,none,none,thv,none,none")]) (define_insn "@pred_narrow_<optab><mode>_scalar" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -4008,7 +4025,8 @@ "TARGET_VECTOR" "vn<insn>.w%o4\t%0,%3,%4%p1" [(set_attr "type" "vnshift") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "spec_restriction" "none,none,thv,thv,none,none")]) ;; vncvt.x.x.w (define_insn "@pred_trunc<mode>" @@ -4032,7 +4050,8 @@ (set_attr "vl_op_idx" "4") (set (attr "ta") (symbol_ref "riscv_vector::get_ta(operands[5])")) (set (attr "ma") (symbol_ref "riscv_vector::get_ma(operands[6])")) - (set (attr "avl_type_idx") (const_int 7))]) + (set (attr "avl_type_idx") (const_int 7)) + (set_attr "spec_restriction" "none,none,thv,thv,none,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated fixed-point operations @@ -4438,7 +4457,8 @@ "TARGET_VECTOR" "vnclip<v_su>.w%o4\t%0,%3,%v4%p1" [(set_attr "type" "vnclip") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "spec_restriction" "thv,thv,thv,thv,thv,thv,none,none,thv,thv,none,none")]) (define_insn "@pred_narrow_clip<v_su><mode>_scalar" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -4460,7 +4480,8 @@ "TARGET_VECTOR" "vnclip<v_su>.w%o4\t%0,%3,%4%p1" [(set_attr "type" "vnclip") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "spec_restriction" "thv,thv,thv,thv,none,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated integer comparison operations @@ -4511,23 +4532,24 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr, &vr, &vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") - (match_operand 7 "const_int_operand" " i, i, i, i") - (match_operand 8 "const_int_operand" " i, i, i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK, rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i, i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_ltge_operator" - [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr") - (match_operand:V_VLSI 5 "vector_arith_operand" " vr, vr, vi, vi")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] + [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr, vr, vr, vr, vr") + (match_operand:V_VLSI 5 "vector_arith_operand" " vr, vr, vi, vi, vr, vr, vi, vi")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0, vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,thv,thv,rvv,rvv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_narrow" @@ -4547,7 +4569,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,thv,thv,thv,thv,none,none")]) (define_expand "@pred_ltge<mode>" [(set (match_operand:<VM> 0 "register_operand") @@ -4591,23 +4614,24 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_ltge<mode>" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr, &vr, &vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") - (match_operand 7 "const_int_operand" " i, i, i, i") - (match_operand 8 "const_int_operand" " i, i, i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK, rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i, i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "ltge_operator" - [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr") - (match_operand:V_VLSI 5 "vector_neg_arith_operand" " vr, vr, vj, vj")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] + [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr, vr, vr, vr, vr") + (match_operand:V_VLSI 5 "vector_neg_arith_operand" " vr, vr, vj, vj, vr, vr, vj, vj")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0, vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,thv,thv,rvv,rvv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_ltge<mode>_narrow" @@ -4627,7 +4651,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,thv,thv,thv,thv,none,none")]) (define_expand "@pred_cmp<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -4673,24 +4698,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_eqge_operator" - [(match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr") + [(match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSI_QHS - (match_operand:<VEL> 5 "register_operand" " r, r"))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r"))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_scalar_narrow" @@ -4711,7 +4737,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,none,none")]) (define_expand "@pred_eqne<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -4757,24 +4784,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSI_QHS - (match_operand:<VEL> 5 "register_operand" " r, r")) - (match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r")) + (match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_eqne<mode>_scalar_narrow" @@ -4795,7 +4823,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,none,none")]) ;; Handle GET_MODE_INNER (mode) = DImode. We need to split them since ;; we need to deal with SEW = 64 in RV32 system. @@ -4922,24 +4951,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_eqge_operator" - [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr") + [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSI_D - (match_operand:<VEL> 5 "register_operand" " r, r"))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r"))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_scalar_narrow" @@ -4960,28 +4990,30 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,none,none")]) ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSI_D - (match_operand:<VEL> 5 "register_operand" " r, r")) - (match_operand:V_VLSI_D 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r")) + (match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_eqne<mode>_scalar_narrow" @@ -5002,7 +5034,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,none,none")]) (define_insn "*pred_cmp<mode>_extended_scalar_merge_tie_mask" [(set (match_operand:<VM> 0 "register_operand" "=vm") @@ -5031,25 +5064,26 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_extended_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_eqge_operator" - [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr") + [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSI_D (sign_extend:<VEL> - (match_operand:<VSUBEL> 5 "register_operand" " r, r")))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VSUBEL> 5 "register_operand" " r, r, r, r")))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) (define_insn "*pred_cmp<mode>_extended_scalar_narrow" [(set (match_operand:<VM> 0 "register_operand" "=vm, vr, vr, &vr, &vr") @@ -5070,7 +5104,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,none,none")]) (define_insn "*pred_eqne<mode>_extended_scalar_merge_tie_mask" [(set (match_operand:<VM> 0 "register_operand" "=vm") @@ -5099,25 +5134,26 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_extended_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSI_D (sign_extend:<VEL> - (match_operand:<VSUBEL> 5 "register_operand" " r, r"))) - (match_operand:V_VLSI_D 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VSUBEL> 5 "register_operand" " r, r, r, r"))) + (match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) (define_insn "*pred_eqne<mode>_extended_scalar_narrow" [(set (match_operand:<VM> 0 "register_operand" "=vm, vr, vr, &vr, &vr") @@ -5138,7 +5174,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,none,none")]) ;; GE, vmsge.vx/vmsgeu.vx ;; @@ -7327,23 +7364,24 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "signed_order_operator" - [(match_operand:V_VLSF 4 "register_operand" " vr, vr") - (match_operand:V_VLSF 5 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + [(match_operand:V_VLSF 4 "register_operand" " vr, vr, vr, vr") + (match_operand:V_VLSF 5 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vmf%B3.vv\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) (define_insn "*pred_cmp<mode>_narrow_merge_tie_mask" [(set (match_operand:<VM> 0 "register_operand" "=vm") @@ -7386,7 +7424,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vmf%B3.vv\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,thv,thv,thv,thv,none,none")]) (define_expand "@pred_cmp<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -7432,24 +7471,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "signed_order_operator" - [(match_operand:V_VLSF 4 "register_operand" " vr, vr") + [(match_operand:V_VLSF 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSF - (match_operand:<VEL> 5 "register_operand" " f, f"))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " f, f, f, f"))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_scalar_narrow" @@ -7470,7 +7510,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,none,none")]) (define_expand "@pred_eqne<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -7516,24 +7557,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSF - (match_operand:<VEL> 5 "register_operand" " f, f")) - (match_operand:V_VLSF 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " f, f, f, f")) + (match_operand:V_VLSF 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_eqne<mode>_scalar_narrow" @@ -7554,7 +7596,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,none,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated floating-point merge @@ -7774,7 +7817,8 @@ [(set_attr "type" "vfncvtftoi") (set_attr "mode" "<VNCONVERT>") (set (attr "frm_mode") - (symbol_ref "riscv_vector::get_frm_mode (operands[8])"))]) + (symbol_ref "riscv_vector::get_frm_mode (operands[8])")) + (set_attr "spec_restriction" "none,none,thv,thv,none,none")]) (define_insn "@pred_narrow_<fix_cvt><mode>" [(set (match_operand:<VNCONVERT> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -7816,7 +7860,8 @@ [(set_attr "type" "vfncvtitof") (set_attr "mode" "<VNCONVERT>") (set (attr "frm_mode") - (symbol_ref "riscv_vector::get_frm_mode (operands[8])"))]) + (symbol_ref "riscv_vector::get_frm_mode (operands[8])")) + (set_attr "spec_restriction" "none,none,thv,thv,none,none")]) (define_insn "@pred_trunc<mode>" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -7839,7 +7884,8 @@ [(set_attr "type" "vfncvtftof") (set_attr "mode" "<V_DOUBLE_TRUNC>") (set (attr "frm_mode") - (symbol_ref "riscv_vector::get_frm_mode (operands[8])"))]) + (symbol_ref "riscv_vector::get_frm_mode (operands[8])")) + (set_attr "spec_restriction" "none,none,thv,thv,none,none")]) (define_insn "@pred_rod_trunc<mode>" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: [PATCH v5] RISC-V: Fix register overlap issue for some xtheadvector instructions 2024-01-11 2:39 ` Jun Sha (Joshua) @ 2024-01-11 2:46 ` juzhe.zhong 2024-01-11 8:12 ` Robin Dapp 0 siblings, 1 reply; 133+ messages in thread From: juzhe.zhong @ 2024-01-11 2:46 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, cooper.joshua, jinma, cooper.qu, Robin Dapp [-- Attachment #1: Type: text/plain, Size: 38629 bytes --] Ok from myside. CCing Robin to see whether he has any more concerns. Thanks. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-11 10:39 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Fix register overlap issue for some xtheadvector instructions For th.vmadc/th.vmsbc as well as narrowing arithmetic instructions and floating-point compare instructions, an illegal instruction exception will be raised if the destination vector register overlaps a source vector register group. To handle this issue, we add an attribute "spec_restriction" to disable some alternatives for xtheadvector. gcc/ChangeLog: * config/riscv/riscv.md (none,thv,rvv): (no,yes): Add an attribute to disable alternative for xtheadvector or RVV1.0. * config/riscv/vector.md: Disable alternatives that destination register overlaps source register group for xtheadvector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config/riscv/riscv.md | 22 +++ gcc/config/riscv/vector.md | 314 +++++++++++++++++++++---------------- 2 files changed, 202 insertions(+), 134 deletions(-) diff --git a/gcc/config/riscv/riscv.md b/gcc/config/riscv/riscv.md index 84212430dc0..23fc32d5cb2 100644 --- a/gcc/config/riscv/riscv.md +++ b/gcc/config/riscv/riscv.md @@ -579,6 +579,25 @@ ] (const_string "yes"))) +;; This attribute marks the alternatives not matching the constraints +;; described in spec as disabled. +(define_attr "spec_restriction" "none,thv,rvv" + (const_string "none")) + +(define_attr "spec_restriction_disabled" "no,yes" + (cond [(eq_attr "spec_restriction" "none") + (const_string "no") + + (and (eq_attr "spec_restriction" "thv") + (match_test "TARGET_XTHEADVECTOR")) + (const_string "yes") + + (and (eq_attr "spec_restriction" "rvv") + (match_test "TARGET_VECTOR && !TARGET_XTHEADVECTOR")) + (const_string "yes") + ] + (const_string "no"))) + ;; Attribute to control enable or disable instructions. (define_attr "enabled" "no,yes" (cond [ @@ -590,6 +609,9 @@ (eq_attr "group_overlap_valid" "no") (const_string "no") + + (eq_attr "spec_restriction_disabled" "yes") + (const_string "no") ] (const_string "yes"))) diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index 3eb6daafbc2..c79416cf0d3 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -3260,7 +3260,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "spec_restriction" "thv,none,none")]) (define_insn "@pred_msbc<mode>" [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr") @@ -3279,7 +3280,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "spec_restriction" "thv,thv,none")]) (define_insn "@pred_madc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3299,7 +3301,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "spec_restriction" "thv,none")]) (define_insn "@pred_msbc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3319,7 +3322,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "spec_restriction" "thv,none")]) (define_expand "@pred_madc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3368,7 +3372,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "spec_restriction" "thv,none")]) (define_insn "*pred_madc<mode>_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3389,7 +3394,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "spec_restriction" "thv,none")]) (define_expand "@pred_msbc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3438,7 +3444,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "spec_restriction" "thv,none")]) (define_insn "*pred_msbc<mode>_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3459,7 +3466,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "spec_restriction" "thv,none")]) (define_insn "@pred_madc<mode>_overflow" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr, &vr") @@ -3477,7 +3485,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "spec_restriction" "thv,none,none")]) (define_insn "@pred_msbc<mode>_overflow" [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") @@ -3495,7 +3504,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "spec_restriction" "thv,thv,none,none")]) (define_insn "@pred_madc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3514,7 +3524,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "spec_restriction" "thv,none")]) (define_insn "@pred_msbc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3533,7 +3544,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "spec_restriction" "thv,none")]) (define_expand "@pred_madc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3580,7 +3592,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "spec_restriction" "thv,none")]) (define_insn "*pred_madc<mode>_overflow_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3600,7 +3613,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "spec_restriction" "thv,none")]) (define_expand "@pred_msbc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3647,7 +3661,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "spec_restriction" "thv,none")]) (define_insn "*pred_msbc<mode>_overflow_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3667,7 +3682,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "spec_restriction" "thv,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated integer unary operations @@ -3987,7 +4003,8 @@ "TARGET_VECTOR" "vn<insn>.w%o4\t%0,%3,%v4%p1" [(set_attr "type" "vnshift") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "spec_restriction" "none,none,thv,thv,none,thv,none,none,none,thv,none,none")]) (define_insn "@pred_narrow_<optab><mode>_scalar" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -4008,7 +4025,8 @@ "TARGET_VECTOR" "vn<insn>.w%o4\t%0,%3,%4%p1" [(set_attr "type" "vnshift") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "spec_restriction" "none,none,thv,thv,none,none")]) ;; vncvt.x.x.w (define_insn "@pred_trunc<mode>" @@ -4032,7 +4050,8 @@ (set_attr "vl_op_idx" "4") (set (attr "ta") (symbol_ref "riscv_vector::get_ta(operands[5])")) (set (attr "ma") (symbol_ref "riscv_vector::get_ma(operands[6])")) - (set (attr "avl_type_idx") (const_int 7))]) + (set (attr "avl_type_idx") (const_int 7)) + (set_attr "spec_restriction" "none,none,thv,thv,none,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated fixed-point operations @@ -4438,7 +4457,8 @@ "TARGET_VECTOR" "vnclip<v_su>.w%o4\t%0,%3,%v4%p1" [(set_attr "type" "vnclip") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "spec_restriction" "thv,thv,thv,thv,thv,thv,none,none,thv,thv,none,none")]) (define_insn "@pred_narrow_clip<v_su><mode>_scalar" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -4460,7 +4480,8 @@ "TARGET_VECTOR" "vnclip<v_su>.w%o4\t%0,%3,%4%p1" [(set_attr "type" "vnclip") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "spec_restriction" "thv,thv,thv,thv,none,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated integer comparison operations @@ -4511,23 +4532,24 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr, &vr, &vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") - (match_operand 7 "const_int_operand" " i, i, i, i") - (match_operand 8 "const_int_operand" " i, i, i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK, rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i, i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_ltge_operator" - [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr") - (match_operand:V_VLSI 5 "vector_arith_operand" " vr, vr, vi, vi")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] + [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr, vr, vr, vr, vr") + (match_operand:V_VLSI 5 "vector_arith_operand" " vr, vr, vi, vi, vr, vr, vi, vi")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0, vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,thv,thv,rvv,rvv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_narrow" @@ -4547,7 +4569,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,thv,thv,thv,thv,none,none")]) (define_expand "@pred_ltge<mode>" [(set (match_operand:<VM> 0 "register_operand") @@ -4591,23 +4614,24 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_ltge<mode>" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr, &vr, &vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") - (match_operand 7 "const_int_operand" " i, i, i, i") - (match_operand 8 "const_int_operand" " i, i, i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK, rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i, i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "ltge_operator" - [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr") - (match_operand:V_VLSI 5 "vector_neg_arith_operand" " vr, vr, vj, vj")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] + [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr, vr, vr, vr, vr") + (match_operand:V_VLSI 5 "vector_neg_arith_operand" " vr, vr, vj, vj, vr, vr, vj, vj")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0, vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,thv,thv,rvv,rvv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_ltge<mode>_narrow" @@ -4627,7 +4651,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,thv,thv,thv,thv,none,none")]) (define_expand "@pred_cmp<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -4673,24 +4698,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_eqge_operator" - [(match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr") + [(match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSI_QHS - (match_operand:<VEL> 5 "register_operand" " r, r"))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r"))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_scalar_narrow" @@ -4711,7 +4737,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,none,none")]) (define_expand "@pred_eqne<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -4757,24 +4784,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSI_QHS - (match_operand:<VEL> 5 "register_operand" " r, r")) - (match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r")) + (match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_eqne<mode>_scalar_narrow" @@ -4795,7 +4823,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,none,none")]) ;; Handle GET_MODE_INNER (mode) = DImode. We need to split them since ;; we need to deal with SEW = 64 in RV32 system. @@ -4922,24 +4951,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_eqge_operator" - [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr") + [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSI_D - (match_operand:<VEL> 5 "register_operand" " r, r"))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r"))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_scalar_narrow" @@ -4960,28 +4990,30 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,none,none")]) ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSI_D - (match_operand:<VEL> 5 "register_operand" " r, r")) - (match_operand:V_VLSI_D 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r")) + (match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_eqne<mode>_scalar_narrow" @@ -5002,7 +5034,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,none,none")]) (define_insn "*pred_cmp<mode>_extended_scalar_merge_tie_mask" [(set (match_operand:<VM> 0 "register_operand" "=vm") @@ -5031,25 +5064,26 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_extended_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_eqge_operator" - [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr") + [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSI_D (sign_extend:<VEL> - (match_operand:<VSUBEL> 5 "register_operand" " r, r")))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VSUBEL> 5 "register_operand" " r, r, r, r")))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) (define_insn "*pred_cmp<mode>_extended_scalar_narrow" [(set (match_operand:<VM> 0 "register_operand" "=vm, vr, vr, &vr, &vr") @@ -5070,7 +5104,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,none,none")]) (define_insn "*pred_eqne<mode>_extended_scalar_merge_tie_mask" [(set (match_operand:<VM> 0 "register_operand" "=vm") @@ -5099,25 +5134,26 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_extended_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSI_D (sign_extend:<VEL> - (match_operand:<VSUBEL> 5 "register_operand" " r, r"))) - (match_operand:V_VLSI_D 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VSUBEL> 5 "register_operand" " r, r, r, r"))) + (match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) (define_insn "*pred_eqne<mode>_extended_scalar_narrow" [(set (match_operand:<VM> 0 "register_operand" "=vm, vr, vr, &vr, &vr") @@ -5138,7 +5174,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,none,none")]) ;; GE, vmsge.vx/vmsgeu.vx ;; @@ -7327,23 +7364,24 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "signed_order_operator" - [(match_operand:V_VLSF 4 "register_operand" " vr, vr") - (match_operand:V_VLSF 5 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + [(match_operand:V_VLSF 4 "register_operand" " vr, vr, vr, vr") + (match_operand:V_VLSF 5 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vmf%B3.vv\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) (define_insn "*pred_cmp<mode>_narrow_merge_tie_mask" [(set (match_operand:<VM> 0 "register_operand" "=vm") @@ -7386,7 +7424,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vmf%B3.vv\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,thv,thv,thv,thv,none,none")]) (define_expand "@pred_cmp<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -7432,24 +7471,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "signed_order_operator" - [(match_operand:V_VLSF 4 "register_operand" " vr, vr") + [(match_operand:V_VLSF 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSF - (match_operand:<VEL> 5 "register_operand" " f, f"))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " f, f, f, f"))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_scalar_narrow" @@ -7470,7 +7510,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,none,none")]) (define_expand "@pred_eqne<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -7516,24 +7557,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSF - (match_operand:<VEL> 5 "register_operand" " f, f")) - (match_operand:V_VLSF 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " f, f, f, f")) + (match_operand:V_VLSF 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_eqne<mode>_scalar_narrow" @@ -7554,7 +7596,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,none,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated floating-point merge @@ -7774,7 +7817,8 @@ [(set_attr "type" "vfncvtftoi") (set_attr "mode" "<VNCONVERT>") (set (attr "frm_mode") - (symbol_ref "riscv_vector::get_frm_mode (operands[8])"))]) + (symbol_ref "riscv_vector::get_frm_mode (operands[8])")) + (set_attr "spec_restriction" "none,none,thv,thv,none,none")]) (define_insn "@pred_narrow_<fix_cvt><mode>" [(set (match_operand:<VNCONVERT> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -7816,7 +7860,8 @@ [(set_attr "type" "vfncvtitof") (set_attr "mode" "<VNCONVERT>") (set (attr "frm_mode") - (symbol_ref "riscv_vector::get_frm_mode (operands[8])"))]) + (symbol_ref "riscv_vector::get_frm_mode (operands[8])")) + (set_attr "spec_restriction" "none,none,thv,thv,none,none")]) (define_insn "@pred_trunc<mode>" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -7839,7 +7884,8 @@ [(set_attr "type" "vfncvtftof") (set_attr "mode" "<V_DOUBLE_TRUNC>") (set (attr "frm_mode") - (symbol_ref "riscv_vector::get_frm_mode (operands[8])"))]) + (symbol_ref "riscv_vector::get_frm_mode (operands[8])")) + (set_attr "spec_restriction" "none,none,thv,thv,none,none")]) (define_insn "@pred_rod_trunc<mode>" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: [PATCH v5] RISC-V: Fix register overlap issue for some xtheadvector instructions 2024-01-11 2:46 ` juzhe.zhong @ 2024-01-11 8:12 ` Robin Dapp 0 siblings, 0 replies; 133+ messages in thread From: Robin Dapp @ 2024-01-11 8:12 UTC (permalink / raw) To: juzhe.zhong, cooper.joshua, gcc-patches Cc: rdapp.gcc, Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu LGTM now, thanks. I find it much more readable that way. Regards Robin ^ permalink raw reply [flat|nested] 133+ messages in thread
* [PATCH v4] RISC-V: Rewrite some instructions using ASM targethook 2023-12-29 4:03 [PATCH v4] RISC-V: Support XTheadVector extension Jun Sha (Joshua) ` (6 preceding siblings ...) 2024-01-03 2:37 ` [PATCH v4] RISC-V: Fix register overlap issue for some xtheadvector instructions Jun Sha (Joshua) @ 2024-01-03 2:39 ` Jun Sha (Joshua) 2024-01-03 3:05 ` Kito Cheng 2024-01-03 7:55 ` Jun Sha (Joshua) 2024-01-12 3:18 ` [PATCH v5] RISC-V: Support XTheadVector extension Jun Sha (Joshua) 8 siblings, 2 replies; 133+ messages in thread From: Jun Sha (Joshua) @ 2024-01-03 2:39 UTC (permalink / raw) To: gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, Jun Sha (Joshua), Jin Ma, Xianmiao Qu There are some xtheadvector instructions that differ from RVV1.0 apart from simply adding "th." prefix. For example, RVV1.0 load/store instructions will have SEW while xtheadvector not; RVV1.0 will have "o" for indexed-ordered store instructions while xtheadvecotr not; xtheadvector and RVV1.0 have different vnsrl/vnsra/vfncvt suffix (vv/vx/vi vs wv/wx/wi). To address this issue without duplicating patterns, we use ASM targethook to rewrite the whole string of the instructions. We identify different instructions from the corresponding attribute. gcc/ChangeLog: * config/riscv/riscv.cc (riscv_asm_output_opcode): Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config/riscv/riscv.cc | 213 +++++++++++++++++++++++++++++++++++++- 1 file changed, 210 insertions(+), 3 deletions(-) diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index a80bf8d1a74..13cdfc4ee27 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -5646,9 +5646,216 @@ riscv_asm_output_opcode (FILE *asm_out_file, const char *p) { /* We need to add th. prefix to all the xtheadvector insturctions here.*/ - if (TARGET_XTHEADVECTOR && current_output_insn != NULL_RTX && - p[0] == 'v') - fputs ("th.", asm_out_file); + if (TARGET_XTHEADVECTOR && current_output_insn != NULL_RTX) + { + if (get_attr_type (current_output_insn) == TYPE_VLDE || + get_attr_type (current_output_insn) == TYPE_VSTE || + get_attr_type (current_output_insn) == TYPE_VLDFF) + { + if (strstr (p, "e8") || strstr (p, "e16") || + strstr (p, "e32") || strstr (p, "e64")) + { + get_attr_type (current_output_insn) == TYPE_VSTE + ? fputs ("th.vse", asm_out_file) + : fputs ("th.vle", asm_out_file); + if (strstr (p, "e8")) + return p+4; + else + return p+5; + } + } + + if (get_attr_type (current_output_insn) == TYPE_VLDS || + get_attr_type (current_output_insn) == TYPE_VSTS) + { + if (strstr (p, "vle8") || strstr (p, "vse8") || + strstr (p, "vle16") || strstr (p, "vse16") || + strstr (p, "vle32") || strstr (p, "vse32") || + strstr (p, "vle64") || strstr (p, "vse64")) + { + get_attr_type (current_output_insn) == TYPE_VSTS + ? fputs ("th.vse", asm_out_file) + : fputs ("th.vle", asm_out_file); + if (strstr (p, "e8")) + return p+4; + else + return p+5; + } + else if (strstr (p, "vlse8") || strstr (p, "vsse8") || + strstr (p, "vlse16") || strstr (p, "vsse16") || + strstr (p, "vlse32") || strstr (p, "vsse32") || + strstr (p, "vlse64") || strstr (p, "vsse64")) + { + get_attr_type (current_output_insn) == TYPE_VSTS + ? fputs ("th.vsse", asm_out_file) + : fputs ("th.vlse", asm_out_file); + if (strstr (p, "e8")) + return p+5; + else + return p+6; + } + } + + if (get_attr_type (current_output_insn) == TYPE_VLDUX || + get_attr_type (current_output_insn) == TYPE_VLDOX) + { + if (strstr (p, "ei")) + { + fputs ("th.vlxe", asm_out_file); + if (strstr (p, "ei8")) + return p+7; + else + return p+8; + } + } + + if (get_attr_type (current_output_insn) == TYPE_VSTUX || + get_attr_type (current_output_insn) == TYPE_VSTOX) + { + if (strstr (p, "ei")) + { + get_attr_type (current_output_insn) == TYPE_VSTUX + ? fputs ("th.vsuxe", asm_out_file) + : fputs ("th.vsxe", asm_out_file); + if (strstr (p, "ei8")) + return p+7; + else + return p+8; + } + } + + if (get_attr_type (current_output_insn) == TYPE_VLSEGDE || + get_attr_type (current_output_insn) == TYPE_VSSEGTE || + get_attr_type (current_output_insn) == TYPE_VLSEGDFF) + { + get_attr_type (current_output_insn) == TYPE_VSSEGTE + ? fputs ("th.vsseg", asm_out_file) + : fputs ("th.vlseg", asm_out_file); + asm_fprintf (asm_out_file, "%c", p[5]); + fputs ("e", asm_out_file); + if (strstr (p, "e8")) + return p+8; + else + return p+9; + } + + if (get_attr_type (current_output_insn) == TYPE_VLSEGDS || + get_attr_type (current_output_insn) == TYPE_VSSEGTS) + { + get_attr_type (current_output_insn) == TYPE_VSSEGTS + ? fputs ("th.vssseg", asm_out_file) + : fputs ("th.vlsseg", asm_out_file); + asm_fprintf (asm_out_file, "%c", p[6]); + fputs ("e", asm_out_file); + if (strstr (p, "e8")) + return p+9; + else + return p+10; + } + + if (get_attr_type (current_output_insn) == TYPE_VLSEGDUX || + get_attr_type (current_output_insn) == TYPE_VLSEGDOX) + { + fputs ("th.vlxseg", asm_out_file); + asm_fprintf (asm_out_file, "%c", p[7]); + fputs ("e", asm_out_file); + if (strstr (p, "e8")) + return p+10; + else + return p+11; + } + + if (get_attr_type (current_output_insn) == TYPE_VSSEGTUX || + get_attr_type (current_output_insn) == TYPE_VSSEGTOX) + { + fputs ("th.vsxseg", asm_out_file); + asm_fprintf (asm_out_file, "%c", p[7]); + fputs ("e", asm_out_file); + if (strstr (p, "e8")) + return p+10; + else + return p+11; + } + + if (get_attr_type (current_output_insn) == TYPE_VNSHIFT) + { + if (strstr (p, "vncvt")) + { + fputs ("th.vncvt.x.x.v", asm_out_file); + return p+11; + } + + strstr (p, "vnsrl") ? fputs ("th.vnsrl.v", asm_out_file) + : fputs ("th.vnsra.v", asm_out_file); + return p+7; + } + + if (get_attr_type (current_output_insn) == TYPE_VNCLIP) + { + if (strstr (p, "vnclipu")) + { + fputs ("th.vnclipu.v", asm_out_file); + return p+9; + } + else + { + fputs ("th.vnclip.v", asm_out_file); + return p+8; + } + } + + if (get_attr_type (current_output_insn) == TYPE_VMPOP) + { + fputs ("th.vmpopc", asm_out_file); + return p+5; + } + + if (get_attr_type (current_output_insn) == TYPE_VMFFS) + { + fputs ("th.vmfirst", asm_out_file); + return p+6; + } + + if (get_attr_type (current_output_insn) == TYPE_VFNCVTFTOI || + get_attr_type (current_output_insn) == TYPE_VFNCVTITOF) + { + if (strstr (p, "xu")) + { + get_attr_type (current_output_insn) == TYPE_VFNCVTFTOI + ? fputs ("th.vfncvt.xu.f.v", asm_out_file) + : fputs ("th.vfncvt.f.xu.v", asm_out_file); + return p+13; + } + else + { + get_attr_type (current_output_insn) == TYPE_VFNCVTFTOI + ? fputs ("th.vfncvt.x.f.v", asm_out_file) + : fputs ("th.vfncvt.f.x.v", asm_out_file); + return p+12; + } + } + + if (get_attr_type (current_output_insn) == TYPE_VFNCVTFTOF) + { + fputs ("th.vfncvt.f.f.v", asm_out_file); + return p+12; + } + + if (get_attr_type (current_output_insn) == TYPE_VFREDU) + { + fputs ("th.vfredsum", asm_out_file); + return p+9; + } + + if (get_attr_type (current_output_insn) == TYPE_VFWREDU) + { + fputs ("th.vfwredsum", asm_out_file); + return p+10; + } + + if (p[0] == 'v') + fputs ("th.", asm_out_file); + } return p; } -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: [PATCH v4] RISC-V: Rewrite some instructions using ASM targethook 2024-01-03 2:39 ` [PATCH v4] RISC-V: Rewrite some instructions using ASM targethook Jun Sha (Joshua) @ 2024-01-03 3:05 ` Kito Cheng 2024-01-03 7:55 ` Jun Sha (Joshua) 1 sibling, 0 replies; 133+ messages in thread From: Kito Cheng @ 2024-01-03 3:05 UTC (permalink / raw) To: Jun Sha (Joshua) Cc: gcc-patches, jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, Jin Ma, Xianmiao Qu Please move those code logic to thead.cc, e.g. if (TARGET_XTHEADVECTOR) return th_asm_output_opcode (asm_out_file, p); And then implement th_asm_output_opcode in thead.cc. On Wed, Jan 3, 2024 at 10:39 AM Jun Sha (Joshua) <cooper.joshua@linux.alibaba.com> wrote: > > There are some xtheadvector instructions that differ from RVV1.0 > apart from simply adding "th." prefix. For example, RVV1.0 > load/store instructions will have SEW while xtheadvector not; > RVV1.0 will have "o" for indexed-ordered store instructions while > xtheadvecotr not; xtheadvector and RVV1.0 have different > vnsrl/vnsra/vfncvt suffix (vv/vx/vi vs wv/wx/wi). > > To address this issue without duplicating patterns, we use ASM > targethook to rewrite the whole string of the instructions. We > identify different instructions from the corresponding attribute. > > gcc/ChangeLog: > > * config/riscv/riscv.cc (riscv_asm_output_opcode): > > Co-authored-by: Jin Ma <jinma@linux.alibaba.com> > Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> > Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> > --- > gcc/config/riscv/riscv.cc | 213 +++++++++++++++++++++++++++++++++++++- > 1 file changed, 210 insertions(+), 3 deletions(-) > > diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc > index a80bf8d1a74..13cdfc4ee27 100644 > --- a/gcc/config/riscv/riscv.cc > +++ b/gcc/config/riscv/riscv.cc > @@ -5646,9 +5646,216 @@ riscv_asm_output_opcode (FILE *asm_out_file, const char *p) > { > /* We need to add th. prefix to all the xtheadvector > insturctions here.*/ > - if (TARGET_XTHEADVECTOR && current_output_insn != NULL_RTX && > - p[0] == 'v') > - fputs ("th.", asm_out_file); > + if (TARGET_XTHEADVECTOR && current_output_insn != NULL_RTX) > + { > + if (get_attr_type (current_output_insn) == TYPE_VLDE || > + get_attr_type (current_output_insn) == TYPE_VSTE || > + get_attr_type (current_output_insn) == TYPE_VLDFF) > + { > + if (strstr (p, "e8") || strstr (p, "e16") || > + strstr (p, "e32") || strstr (p, "e64")) > + { > + get_attr_type (current_output_insn) == TYPE_VSTE > + ? fputs ("th.vse", asm_out_file) > + : fputs ("th.vle", asm_out_file); > + if (strstr (p, "e8")) > + return p+4; > + else > + return p+5; > + } > + } > + > + if (get_attr_type (current_output_insn) == TYPE_VLDS || > + get_attr_type (current_output_insn) == TYPE_VSTS) > + { > + if (strstr (p, "vle8") || strstr (p, "vse8") || > + strstr (p, "vle16") || strstr (p, "vse16") || > + strstr (p, "vle32") || strstr (p, "vse32") || > + strstr (p, "vle64") || strstr (p, "vse64")) > + { > + get_attr_type (current_output_insn) == TYPE_VSTS > + ? fputs ("th.vse", asm_out_file) > + : fputs ("th.vle", asm_out_file); > + if (strstr (p, "e8")) > + return p+4; > + else > + return p+5; > + } > + else if (strstr (p, "vlse8") || strstr (p, "vsse8") || > + strstr (p, "vlse16") || strstr (p, "vsse16") || > + strstr (p, "vlse32") || strstr (p, "vsse32") || > + strstr (p, "vlse64") || strstr (p, "vsse64")) > + { > + get_attr_type (current_output_insn) == TYPE_VSTS > + ? fputs ("th.vsse", asm_out_file) > + : fputs ("th.vlse", asm_out_file); > + if (strstr (p, "e8")) > + return p+5; > + else > + return p+6; > + } > + } > + > + if (get_attr_type (current_output_insn) == TYPE_VLDUX || > + get_attr_type (current_output_insn) == TYPE_VLDOX) > + { > + if (strstr (p, "ei")) > + { > + fputs ("th.vlxe", asm_out_file); > + if (strstr (p, "ei8")) > + return p+7; > + else > + return p+8; > + } > + } > + > + if (get_attr_type (current_output_insn) == TYPE_VSTUX || > + get_attr_type (current_output_insn) == TYPE_VSTOX) > + { > + if (strstr (p, "ei")) > + { > + get_attr_type (current_output_insn) == TYPE_VSTUX > + ? fputs ("th.vsuxe", asm_out_file) > + : fputs ("th.vsxe", asm_out_file); > + if (strstr (p, "ei8")) > + return p+7; > + else > + return p+8; > + } > + } > + > + if (get_attr_type (current_output_insn) == TYPE_VLSEGDE || > + get_attr_type (current_output_insn) == TYPE_VSSEGTE || > + get_attr_type (current_output_insn) == TYPE_VLSEGDFF) > + { > + get_attr_type (current_output_insn) == TYPE_VSSEGTE > + ? fputs ("th.vsseg", asm_out_file) > + : fputs ("th.vlseg", asm_out_file); > + asm_fprintf (asm_out_file, "%c", p[5]); > + fputs ("e", asm_out_file); > + if (strstr (p, "e8")) > + return p+8; > + else > + return p+9; > + } > + > + if (get_attr_type (current_output_insn) == TYPE_VLSEGDS || > + get_attr_type (current_output_insn) == TYPE_VSSEGTS) > + { > + get_attr_type (current_output_insn) == TYPE_VSSEGTS > + ? fputs ("th.vssseg", asm_out_file) > + : fputs ("th.vlsseg", asm_out_file); > + asm_fprintf (asm_out_file, "%c", p[6]); > + fputs ("e", asm_out_file); > + if (strstr (p, "e8")) > + return p+9; > + else > + return p+10; > + } > + > + if (get_attr_type (current_output_insn) == TYPE_VLSEGDUX || > + get_attr_type (current_output_insn) == TYPE_VLSEGDOX) > + { > + fputs ("th.vlxseg", asm_out_file); > + asm_fprintf (asm_out_file, "%c", p[7]); > + fputs ("e", asm_out_file); > + if (strstr (p, "e8")) > + return p+10; > + else > + return p+11; > + } > + > + if (get_attr_type (current_output_insn) == TYPE_VSSEGTUX || > + get_attr_type (current_output_insn) == TYPE_VSSEGTOX) > + { > + fputs ("th.vsxseg", asm_out_file); > + asm_fprintf (asm_out_file, "%c", p[7]); > + fputs ("e", asm_out_file); > + if (strstr (p, "e8")) > + return p+10; > + else > + return p+11; > + } > + > + if (get_attr_type (current_output_insn) == TYPE_VNSHIFT) > + { > + if (strstr (p, "vncvt")) > + { > + fputs ("th.vncvt.x.x.v", asm_out_file); > + return p+11; > + } > + > + strstr (p, "vnsrl") ? fputs ("th.vnsrl.v", asm_out_file) > + : fputs ("th.vnsra.v", asm_out_file); > + return p+7; > + } > + > + if (get_attr_type (current_output_insn) == TYPE_VNCLIP) > + { > + if (strstr (p, "vnclipu")) > + { > + fputs ("th.vnclipu.v", asm_out_file); > + return p+9; > + } > + else > + { > + fputs ("th.vnclip.v", asm_out_file); > + return p+8; > + } > + } > + > + if (get_attr_type (current_output_insn) == TYPE_VMPOP) > + { > + fputs ("th.vmpopc", asm_out_file); > + return p+5; > + } > + > + if (get_attr_type (current_output_insn) == TYPE_VMFFS) > + { > + fputs ("th.vmfirst", asm_out_file); > + return p+6; > + } > + > + if (get_attr_type (current_output_insn) == TYPE_VFNCVTFTOI || > + get_attr_type (current_output_insn) == TYPE_VFNCVTITOF) > + { > + if (strstr (p, "xu")) > + { > + get_attr_type (current_output_insn) == TYPE_VFNCVTFTOI > + ? fputs ("th.vfncvt.xu.f.v", asm_out_file) > + : fputs ("th.vfncvt.f.xu.v", asm_out_file); > + return p+13; > + } > + else > + { > + get_attr_type (current_output_insn) == TYPE_VFNCVTFTOI > + ? fputs ("th.vfncvt.x.f.v", asm_out_file) > + : fputs ("th.vfncvt.f.x.v", asm_out_file); > + return p+12; > + } > + } > + > + if (get_attr_type (current_output_insn) == TYPE_VFNCVTFTOF) > + { > + fputs ("th.vfncvt.f.f.v", asm_out_file); > + return p+12; > + } > + > + if (get_attr_type (current_output_insn) == TYPE_VFREDU) > + { > + fputs ("th.vfredsum", asm_out_file); > + return p+9; > + } > + > + if (get_attr_type (current_output_insn) == TYPE_VFWREDU) > + { > + fputs ("th.vfwredsum", asm_out_file); > + return p+10; > + } > + > + if (p[0] == 'v') > + fputs ("th.", asm_out_file); > + } > > return p; > } > -- > 2.17.1 > ^ permalink raw reply [flat|nested] 133+ messages in thread
* [PATCH v4] RISC-V: Rewrite some instructions using ASM targethook 2024-01-03 2:39 ` [PATCH v4] RISC-V: Rewrite some instructions using ASM targethook Jun Sha (Joshua) 2024-01-03 3:05 ` Kito Cheng @ 2024-01-03 7:55 ` Jun Sha (Joshua) 1 sibling, 0 replies; 133+ messages in thread From: Jun Sha (Joshua) @ 2024-01-03 7:55 UTC (permalink / raw) To: gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, Jun Sha (Joshua), Jin Ma, Xianmiao Qu There are some xtheadvector instructions that differ from RVV1.0 apart from simply adding "th." prefix. For example, RVV1.0 load/store instructions will have SEW while xtheadvector not; RVV1.0 will have "o" for indexed-ordered store instructions while xtheadvecotr not; xtheadvector and RVV1.0 have different vnsrl/vnsra/vfncvt suffix (vv/vx/vi vs wv/wx/wi). To address this issue without duplicating patterns, we use ASM targethook to rewrite the whole string of the instructions. We identify different instructions from the corresponding attribute. gcc/ChangeLog: * config/riscv/thead.cc (th_asm_output_opcode): Rewrite some instructions. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config/riscv/thead.cc | 215 +++++++++++++++++++++++++++++++++++++- 1 file changed, 213 insertions(+), 2 deletions(-) diff --git a/gcc/config/riscv/thead.cc b/gcc/config/riscv/thead.cc index dc3aed3904d..fb088ebff02 100644 --- a/gcc/config/riscv/thead.cc +++ b/gcc/config/riscv/thead.cc @@ -27,6 +27,7 @@ #include "backend.h" #include "tree.h" #include "rtl.h" +#include "insn-attr.h" #include "explow.h" #include "memmodel.h" #include "emit-rtl.h" @@ -890,8 +891,218 @@ th_asm_output_opcode (FILE *asm_out_file, const char *p) { /* We need to add th. prefix to all the xtheadvector instructions here.*/ - if (current_output_insn != NULL && p[0] == 'v') - fputs ("th.", asm_out_file); + if (current_output_insn != NULL) + { + if (get_attr_type (current_output_insn) == TYPE_VLDE || + get_attr_type (current_output_insn) == TYPE_VSTE || + get_attr_type (current_output_insn) == TYPE_VLDFF) + { + if (strstr (p, "e8") || strstr (p, "e16") || + strstr (p, "e32") || strstr (p, "e64")) + { + get_attr_type (current_output_insn) == TYPE_VSTE + ? fputs ("th.vse", asm_out_file) + : fputs ("th.vle", asm_out_file); + if (strstr (p, "e8")) + return p+4; + else + return p+5; + } + } + + if (get_attr_type (current_output_insn) == TYPE_VLDS || + get_attr_type (current_output_insn) == TYPE_VSTS) + { + if (strstr (p, "vle8") || strstr (p, "vse8") || + strstr (p, "vle16") || strstr (p, "vse16") || + strstr (p, "vle32") || strstr (p, "vse32") || + strstr (p, "vle64") || strstr (p, "vse64")) + { + get_attr_type (current_output_insn) == TYPE_VSTS + ? fputs ("th.vse", asm_out_file) + : fputs ("th.vle", asm_out_file); + if (strstr (p, "e8")) + return p+4; + else + return p+5; + } + else if (strstr (p, "vlse8") || strstr (p, "vsse8") || + strstr (p, "vlse16") || strstr (p, "vsse16") || + strstr (p, "vlse32") || strstr (p, "vsse32") || + strstr (p, "vlse64") || strstr (p, "vsse64")) + { + get_attr_type (current_output_insn) == TYPE_VSTS + ? fputs ("th.vsse", asm_out_file) + : fputs ("th.vlse", asm_out_file); + if (strstr (p, "e8")) + return p+5; + else + return p+6; + } + } + + if (get_attr_type (current_output_insn) == TYPE_VLDUX || + get_attr_type (current_output_insn) == TYPE_VLDOX) + { + if (strstr (p, "ei")) + { + fputs ("th.vlxe", asm_out_file); + if (strstr (p, "ei8")) + return p+7; + else + return p+8; + } + } + + if (get_attr_type (current_output_insn) == TYPE_VSTUX || + get_attr_type (current_output_insn) == TYPE_VSTOX) + { + if (strstr (p, "ei")) + { + get_attr_type (current_output_insn) == TYPE_VSTUX + ? fputs ("th.vsuxe", asm_out_file) + : fputs ("th.vsxe", asm_out_file); + if (strstr (p, "ei8")) + return p+7; + else + return p+8; + } + } + + if (get_attr_type (current_output_insn) == TYPE_VLSEGDE || + get_attr_type (current_output_insn) == TYPE_VSSEGTE || + get_attr_type (current_output_insn) == TYPE_VLSEGDFF) + { + get_attr_type (current_output_insn) == TYPE_VSSEGTE + ? fputs ("th.vsseg", asm_out_file) + : fputs ("th.vlseg", asm_out_file); + asm_fprintf (asm_out_file, "%c", p[5]); + fputs ("e", asm_out_file); + if (strstr (p, "e8")) + return p+8; + else + return p+9; + } + + if (get_attr_type (current_output_insn) == TYPE_VLSEGDS || + get_attr_type (current_output_insn) == TYPE_VSSEGTS) + { + get_attr_type (current_output_insn) == TYPE_VSSEGTS + ? fputs ("th.vssseg", asm_out_file) + : fputs ("th.vlsseg", asm_out_file); + asm_fprintf (asm_out_file, "%c", p[6]); + fputs ("e", asm_out_file); + if (strstr (p, "e8")) + return p+9; + else + return p+10; + } + + if (get_attr_type (current_output_insn) == TYPE_VLSEGDUX || + get_attr_type (current_output_insn) == TYPE_VLSEGDOX) + { + fputs ("th.vlxseg", asm_out_file); + asm_fprintf (asm_out_file, "%c", p[7]); + fputs ("e", asm_out_file); + if (strstr (p, "ei8")) + return p+11; + else + return p+12; + } + + if (get_attr_type (current_output_insn) == TYPE_VSSEGTUX || + get_attr_type (current_output_insn) == TYPE_VSSEGTOX) + { + fputs ("th.vsxseg", asm_out_file); + asm_fprintf (asm_out_file, "%c", p[7]); + fputs ("e", asm_out_file); + if (strstr (p, "ei8")) + return p+11; + else + return p+12; + } + + if (get_attr_type (current_output_insn) == TYPE_VNSHIFT) + { + if (strstr (p, "vncvt")) + { + fputs ("th.vncvt.x.x.v", asm_out_file); + return p+11; + } + + strstr (p, "vnsrl") ? fputs ("th.vnsrl.v", asm_out_file) + : fputs ("th.vnsra.v", asm_out_file); + return p+7; + } + + if (get_attr_type (current_output_insn) == TYPE_VNCLIP) + { + if (strstr (p, "vnclipu")) + { + fputs ("th.vnclipu.v", asm_out_file); + return p+9; + } + else + { + fputs ("th.vnclip.v", asm_out_file); + return p+8; + } + } + + if (get_attr_type (current_output_insn) == TYPE_VMPOP) + { + fputs ("th.vmpopc", asm_out_file); + return p+5; + } + + if (get_attr_type (current_output_insn) == TYPE_VMFFS) + { + fputs ("th.vmfirst", asm_out_file); + return p+6; + } + + if (get_attr_type (current_output_insn) == TYPE_VFNCVTFTOI || + get_attr_type (current_output_insn) == TYPE_VFNCVTITOF) + { + if (strstr (p, "xu")) + { + get_attr_type (current_output_insn) == TYPE_VFNCVTFTOI + ? fputs ("th.vfncvt.xu.f.v", asm_out_file) + : fputs ("th.vfncvt.f.xu.v", asm_out_file); + return p+13; + } + else + { + get_attr_type (current_output_insn) == TYPE_VFNCVTFTOI + ? fputs ("th.vfncvt.x.f.v", asm_out_file) + : fputs ("th.vfncvt.f.x.v", asm_out_file); + return p+12; + } + } + + if (get_attr_type (current_output_insn) == TYPE_VFNCVTFTOF) + { + fputs ("th.vfncvt.f.f.v", asm_out_file); + return p+12; + } + + if (get_attr_type (current_output_insn) == TYPE_VFREDU + && strstr (p, "sum")) + { + fputs ("th.vfredsum", asm_out_file); + return p+9; + } + + if (get_attr_type (current_output_insn) == TYPE_VFWREDU + && strstr (p, "sum")) + { + fputs ("th.vfwredsum", asm_out_file); + return p+10; + } + + if (p[0] == 'v') + fputs ("th.", asm_out_file); + } return p; } -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* [PATCH v5] RISC-V: Support XTheadVector extension 2023-12-29 4:03 [PATCH v4] RISC-V: Support XTheadVector extension Jun Sha (Joshua) ` (7 preceding siblings ...) 2024-01-03 2:39 ` [PATCH v4] RISC-V: Rewrite some instructions using ASM targethook Jun Sha (Joshua) @ 2024-01-12 3:18 ` Jun Sha (Joshua) 2024-01-12 3:20 ` [PATCH v4] RISC-V: Introduce XTheadVector as a subset of V1.0.0 Jun Sha (Joshua) ` (6 more replies) 8 siblings, 7 replies; 133+ messages in thread From: Jun Sha (Joshua) @ 2024-01-12 3:18 UTC (permalink / raw) To: gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, kito.cheng, Jun Sha (Joshua), Jin Ma, Xianmiao Qu This patch series presents gcc implementation of the XTheadVector extension [1]. [1] https://github.com/T-head-Semi/thead-extension-spec/ For some vector patterns that cannot be avoided, we use "!TARGET_XTHEADVECTOR" to disable them in order not to generate instructions that xtheadvector does not support, causing 10 changes in vector.md. For the th. prefix issue, we use current_output_insn and the ASM_OUTPUT_OPCODE hook instead of directly modifying patterns in vector.md. We have run the GCC test suite and can confirm that there are no regressions. Furthermore, we have run the tests in https://github.com/riscv-non-isa/rvv-intrinsic-doc/tree/main/examples, and all the tests passed. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> [PATCH v4] RISC-V: Introduce XTheadVector as a subset of V1.0.0 [PATCH v5] RISC-V: Adds the prefix "th." for the instructions of XTheadVector [PATCH v6] RISC-V: Handle differences between XTheadvector and Vector [PATCH v6] RISC-V: Add support for xtheadvector-specific intrinsics [PATCH v6] RISC-V: Fix register overlap issue for some xtheadvector instructions [PATCH v5] RISC-V: Rewrite some instructions using ASM targethook ^ permalink raw reply [flat|nested] 133+ messages in thread
* [PATCH v4] RISC-V: Introduce XTheadVector as a subset of V1.0.0 2024-01-12 3:18 ` [PATCH v5] RISC-V: Support XTheadVector extension Jun Sha (Joshua) @ 2024-01-12 3:20 ` Jun Sha (Joshua) 2024-01-12 7:31 ` juzhe.zhong 2024-01-12 3:21 ` [PATCH v5] RISC-V: Adds the prefix "th." for the instructions of XTheadVector Jun Sha (Joshua) ` (5 subsequent siblings) 6 siblings, 1 reply; 133+ messages in thread From: Jun Sha (Joshua) @ 2024-01-12 3:20 UTC (permalink / raw) To: gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, kito.cheng, Jun Sha (Joshua), Jin Ma, Xianmiao Qu This patch is to introduce basic XTheadVector support (march string parsing and a test for __riscv_xtheadvector) according to https://github.com/T-head-Semi/thead-extension-spec/ gcc/ChangeLog: * common/config/riscv/riscv-common.cc (riscv_subset_list::parse): Add new vendor extension. * config/riscv/riscv-c.cc (riscv_cpu_cpp_builtins): Add test marco. * config/riscv/riscv.opt: Add new mask. gcc/testsuite/ChangeLog: * gcc.target/riscv/predef-__riscv_th_v_intrinsic.c: New test. * gcc.target/riscv/rvv/xtheadvector.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/common/config/riscv/riscv-common.cc | 23 +++++++++++++++++++ gcc/config/riscv/riscv-c.cc | 8 +++++-- gcc/config/riscv/riscv.opt | 2 ++ .../riscv/predef-__riscv_th_v_intrinsic.c | 11 +++++++++ .../gcc.target/riscv/rvv/xtheadvector.c | 13 +++++++++++ 5 files changed, 55 insertions(+), 2 deletions(-) create mode 100644 gcc/testsuite/gcc.target/riscv/predef-__riscv_th_v_intrinsic.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector.c diff --git a/gcc/common/config/riscv/riscv-common.cc b/gcc/common/config/riscv/riscv-common.cc index 0301d170a41..449722070d4 100644 --- a/gcc/common/config/riscv/riscv-common.cc +++ b/gcc/common/config/riscv/riscv-common.cc @@ -368,6 +368,7 @@ static const struct riscv_ext_version riscv_ext_version_table[] = {"xtheadmemidx", ISA_SPEC_CLASS_NONE, 1, 0}, {"xtheadmempair", ISA_SPEC_CLASS_NONE, 1, 0}, {"xtheadsync", ISA_SPEC_CLASS_NONE, 1, 0}, + {"xtheadvector", ISA_SPEC_CLASS_NONE, 1, 0}, {"xventanacondops", ISA_SPEC_CLASS_NONE, 1, 0}, @@ -1251,6 +1252,15 @@ riscv_subset_list::check_conflict_ext () if (lookup ("zcmp")) error_at (m_loc, "%<-march=%s%>: zcd conflicts with zcmp", m_arch); } + + if ((lookup ("v") || lookup ("zve32x") + || lookup ("zve64x") || lookup ("zve32f") + || lookup ("zve64f") || lookup ("zve64d") + || lookup ("zvl32b") || lookup ("zvl64b") + || lookup ("zvl128b") || lookup ("zvfh")) + && lookup ("xtheadvector")) + error_at (m_loc, "%<-march=%s%>: xtheadvector conflicts with vector " + "extension or its sub-extensions", m_arch); } /* Parsing function for multi-letter extensions. @@ -1743,6 +1753,19 @@ static const riscv_ext_flag_table_t riscv_ext_flag_table[] = {"xtheadmemidx", &gcc_options::x_riscv_xthead_subext, MASK_XTHEADMEMIDX}, {"xtheadmempair", &gcc_options::x_riscv_xthead_subext, MASK_XTHEADMEMPAIR}, {"xtheadsync", &gcc_options::x_riscv_xthead_subext, MASK_XTHEADSYNC}, + {"xtheadvector", &gcc_options::x_riscv_xthead_subext, MASK_XTHEADVECTOR}, + {"xtheadvector", &gcc_options::x_riscv_vector_elen_flags, MASK_VECTOR_ELEN_32}, + {"xtheadvector", &gcc_options::x_riscv_vector_elen_flags, MASK_VECTOR_ELEN_64}, + {"xtheadvector", &gcc_options::x_riscv_vector_elen_flags, MASK_VECTOR_ELEN_FP_32}, + {"xtheadvector", &gcc_options::x_riscv_vector_elen_flags, MASK_VECTOR_ELEN_FP_64}, + {"xtheadvector", &gcc_options::x_riscv_vector_elen_flags, MASK_VECTOR_ELEN_FP_16}, + {"xtheadvector", &gcc_options::x_riscv_zvl_flags, MASK_ZVL32B}, + {"xtheadvector", &gcc_options::x_riscv_zvl_flags, MASK_ZVL64B}, + {"xtheadvector", &gcc_options::x_riscv_zvl_flags, MASK_ZVL128B}, + {"xtheadvector", &gcc_options::x_riscv_zf_subext, MASK_ZVFHMIN}, + {"xtheadvector", &gcc_options::x_riscv_zf_subext, MASK_ZVFH}, + {"xtheadvector", &gcc_options::x_target_flags, MASK_FULL_V}, + {"xtheadvector", &gcc_options::x_target_flags, MASK_VECTOR}, {"xventanacondops", &gcc_options::x_riscv_xventana_subext, MASK_XVENTANACONDOPS}, diff --git a/gcc/config/riscv/riscv-c.cc b/gcc/config/riscv/riscv-c.cc index ba60cd8b555..422ddc2c308 100644 --- a/gcc/config/riscv/riscv-c.cc +++ b/gcc/config/riscv/riscv-c.cc @@ -142,6 +142,10 @@ riscv_cpu_cpp_builtins (cpp_reader *pfile) riscv_ext_version_value (0, 11)); } + if (TARGET_XTHEADVECTOR) + builtin_define_with_int_value ("__riscv_th_v_intrinsic", + riscv_ext_version_value (0, 11)); + /* Define architecture extension test macros. */ builtin_define_with_int_value ("__riscv_arch_test", 1); @@ -195,8 +199,8 @@ riscv_pragma_intrinsic (cpp_reader *) { if (!TARGET_VECTOR) { - error ("%<#pragma riscv intrinsic%> option %qs needs 'V' extension " - "enabled", + error ("%<#pragma riscv intrinsic%> option %qs needs 'V' or " + "'XTHEADVECTOR' extension enabled", name); return; } diff --git a/gcc/config/riscv/riscv.opt b/gcc/config/riscv/riscv.opt index 44ed6d69da2..bb18a22b693 100644 --- a/gcc/config/riscv/riscv.opt +++ b/gcc/config/riscv/riscv.opt @@ -452,6 +452,8 @@ Mask(XTHEADMEMPAIR) Var(riscv_xthead_subext) Mask(XTHEADSYNC) Var(riscv_xthead_subext) +Mask(XTHEADVECTOR) Var(riscv_xthead_subext) + TargetVariable int riscv_xventana_subext diff --git a/gcc/testsuite/gcc.target/riscv/predef-__riscv_th_v_intrinsic.c b/gcc/testsuite/gcc.target/riscv/predef-__riscv_th_v_intrinsic.c new file mode 100644 index 00000000000..550b9039a06 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/predef-__riscv_th_v_intrinsic.c @@ -0,0 +1,11 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64imafdcxtheadvector -mabi=lp64d" } */ + +int main () { + +#if __riscv_th_v_intrinsic != 11000 +#error "__riscv_th_v_intrinsic" +#endif + + return 0; +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector.c new file mode 100644 index 00000000000..8ad370172e3 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector.c @@ -0,0 +1,13 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gc_xtheadvector" { target { rv32 } } } */ +/* { dg-options "-march=rv64gc_xtheadvector" { target { rv64 } } } */ + +#ifndef __riscv_xtheadvector +#error "Feature macro not defined" +#endif + +int +foo (int a) +{ + return a; +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: [PATCH v4] RISC-V: Introduce XTheadVector as a subset of V1.0.0 2024-01-12 3:20 ` [PATCH v4] RISC-V: Introduce XTheadVector as a subset of V1.0.0 Jun Sha (Joshua) @ 2024-01-12 7:31 ` juzhe.zhong 2024-01-18 9:50 ` Kito Cheng 0 siblings, 1 reply; 133+ messages in thread From: juzhe.zhong @ 2024-01-12 7:31 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, kito.cheng, cooper.joshua, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 6531 bytes --] This patch needs kito review. I can't approve that. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-12 11:20 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; kito.cheng; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v4] RISC-V: Introduce XTheadVector as a subset of V1.0.0 This patch is to introduce basic XTheadVector support (march string parsing and a test for __riscv_xtheadvector) according to https://github.com/T-head-Semi/thead-extension-spec/ gcc/ChangeLog: * common/config/riscv/riscv-common.cc (riscv_subset_list::parse): Add new vendor extension. * config/riscv/riscv-c.cc (riscv_cpu_cpp_builtins): Add test marco. * config/riscv/riscv.opt: Add new mask. gcc/testsuite/ChangeLog: * gcc.target/riscv/predef-__riscv_th_v_intrinsic.c: New test. * gcc.target/riscv/rvv/xtheadvector.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/common/config/riscv/riscv-common.cc | 23 +++++++++++++++++++ gcc/config/riscv/riscv-c.cc | 8 +++++-- gcc/config/riscv/riscv.opt | 2 ++ .../riscv/predef-__riscv_th_v_intrinsic.c | 11 +++++++++ .../gcc.target/riscv/rvv/xtheadvector.c | 13 +++++++++++ 5 files changed, 55 insertions(+), 2 deletions(-) create mode 100644 gcc/testsuite/gcc.target/riscv/predef-__riscv_th_v_intrinsic.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector.c diff --git a/gcc/common/config/riscv/riscv-common.cc b/gcc/common/config/riscv/riscv-common.cc index 0301d170a41..449722070d4 100644 --- a/gcc/common/config/riscv/riscv-common.cc +++ b/gcc/common/config/riscv/riscv-common.cc @@ -368,6 +368,7 @@ static const struct riscv_ext_version riscv_ext_version_table[] = {"xtheadmemidx", ISA_SPEC_CLASS_NONE, 1, 0}, {"xtheadmempair", ISA_SPEC_CLASS_NONE, 1, 0}, {"xtheadsync", ISA_SPEC_CLASS_NONE, 1, 0}, + {"xtheadvector", ISA_SPEC_CLASS_NONE, 1, 0}, {"xventanacondops", ISA_SPEC_CLASS_NONE, 1, 0}, @@ -1251,6 +1252,15 @@ riscv_subset_list::check_conflict_ext () if (lookup ("zcmp")) error_at (m_loc, "%<-march=%s%>: zcd conflicts with zcmp", m_arch); } + + if ((lookup ("v") || lookup ("zve32x") + || lookup ("zve64x") || lookup ("zve32f") + || lookup ("zve64f") || lookup ("zve64d") + || lookup ("zvl32b") || lookup ("zvl64b") + || lookup ("zvl128b") || lookup ("zvfh")) + && lookup ("xtheadvector")) + error_at (m_loc, "%<-march=%s%>: xtheadvector conflicts with vector " + "extension or its sub-extensions", m_arch); } /* Parsing function for multi-letter extensions. @@ -1743,6 +1753,19 @@ static const riscv_ext_flag_table_t riscv_ext_flag_table[] = {"xtheadmemidx", &gcc_options::x_riscv_xthead_subext, MASK_XTHEADMEMIDX}, {"xtheadmempair", &gcc_options::x_riscv_xthead_subext, MASK_XTHEADMEMPAIR}, {"xtheadsync", &gcc_options::x_riscv_xthead_subext, MASK_XTHEADSYNC}, + {"xtheadvector", &gcc_options::x_riscv_xthead_subext, MASK_XTHEADVECTOR}, + {"xtheadvector", &gcc_options::x_riscv_vector_elen_flags, MASK_VECTOR_ELEN_32}, + {"xtheadvector", &gcc_options::x_riscv_vector_elen_flags, MASK_VECTOR_ELEN_64}, + {"xtheadvector", &gcc_options::x_riscv_vector_elen_flags, MASK_VECTOR_ELEN_FP_32}, + {"xtheadvector", &gcc_options::x_riscv_vector_elen_flags, MASK_VECTOR_ELEN_FP_64}, + {"xtheadvector", &gcc_options::x_riscv_vector_elen_flags, MASK_VECTOR_ELEN_FP_16}, + {"xtheadvector", &gcc_options::x_riscv_zvl_flags, MASK_ZVL32B}, + {"xtheadvector", &gcc_options::x_riscv_zvl_flags, MASK_ZVL64B}, + {"xtheadvector", &gcc_options::x_riscv_zvl_flags, MASK_ZVL128B}, + {"xtheadvector", &gcc_options::x_riscv_zf_subext, MASK_ZVFHMIN}, + {"xtheadvector", &gcc_options::x_riscv_zf_subext, MASK_ZVFH}, + {"xtheadvector", &gcc_options::x_target_flags, MASK_FULL_V}, + {"xtheadvector", &gcc_options::x_target_flags, MASK_VECTOR}, {"xventanacondops", &gcc_options::x_riscv_xventana_subext, MASK_XVENTANACONDOPS}, diff --git a/gcc/config/riscv/riscv-c.cc b/gcc/config/riscv/riscv-c.cc index ba60cd8b555..422ddc2c308 100644 --- a/gcc/config/riscv/riscv-c.cc +++ b/gcc/config/riscv/riscv-c.cc @@ -142,6 +142,10 @@ riscv_cpu_cpp_builtins (cpp_reader *pfile) riscv_ext_version_value (0, 11)); } + if (TARGET_XTHEADVECTOR) + builtin_define_with_int_value ("__riscv_th_v_intrinsic", + riscv_ext_version_value (0, 11)); + /* Define architecture extension test macros. */ builtin_define_with_int_value ("__riscv_arch_test", 1); @@ -195,8 +199,8 @@ riscv_pragma_intrinsic (cpp_reader *) { if (!TARGET_VECTOR) { - error ("%<#pragma riscv intrinsic%> option %qs needs 'V' extension " - "enabled", + error ("%<#pragma riscv intrinsic%> option %qs needs 'V' or " + "'XTHEADVECTOR' extension enabled", name); return; } diff --git a/gcc/config/riscv/riscv.opt b/gcc/config/riscv/riscv.opt index 44ed6d69da2..bb18a22b693 100644 --- a/gcc/config/riscv/riscv.opt +++ b/gcc/config/riscv/riscv.opt @@ -452,6 +452,8 @@ Mask(XTHEADMEMPAIR) Var(riscv_xthead_subext) Mask(XTHEADSYNC) Var(riscv_xthead_subext) +Mask(XTHEADVECTOR) Var(riscv_xthead_subext) + TargetVariable int riscv_xventana_subext diff --git a/gcc/testsuite/gcc.target/riscv/predef-__riscv_th_v_intrinsic.c b/gcc/testsuite/gcc.target/riscv/predef-__riscv_th_v_intrinsic.c new file mode 100644 index 00000000000..550b9039a06 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/predef-__riscv_th_v_intrinsic.c @@ -0,0 +1,11 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv64imafdcxtheadvector -mabi=lp64d" } */ + +int main () { + +#if __riscv_th_v_intrinsic != 11000 +#error "__riscv_th_v_intrinsic" +#endif + + return 0; +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector.c new file mode 100644 index 00000000000..8ad370172e3 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector.c @@ -0,0 +1,13 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gc_xtheadvector" { target { rv32 } } } */ +/* { dg-options "-march=rv64gc_xtheadvector" { target { rv64 } } } */ + +#ifndef __riscv_xtheadvector +#error "Feature macro not defined" +#endif + +int +foo (int a) +{ + return a; +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: [PATCH v4] RISC-V: Introduce XTheadVector as a subset of V1.0.0 2024-01-12 7:31 ` juzhe.zhong @ 2024-01-18 9:50 ` Kito Cheng 0 siblings, 0 replies; 133+ messages in thread From: Kito Cheng @ 2024-01-18 9:50 UTC (permalink / raw) To: juzhe.zhong Cc: cooper.joshua, gcc-patches, Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu LGTM On Fri, Jan 12, 2024 at 3:32 PM juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> wrote: > > This patch needs kito review. I can't approve that. > > ________________________________ > juzhe.zhong@rivai.ai > > > From: Jun Sha (Joshua) > Date: 2024-01-12 11:20 > To: gcc-patches > CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; kito.cheng; Jun Sha (Joshua); Jin Ma; Xianmiao Qu > Subject: [PATCH v4] RISC-V: Introduce XTheadVector as a subset of V1.0.0 > This patch is to introduce basic XTheadVector support > (march string parsing and a test for __riscv_xtheadvector) > according to https://github.com/T-head-Semi/thead-extension-spec/ > > gcc/ChangeLog: > > * common/config/riscv/riscv-common.cc > (riscv_subset_list::parse): Add new vendor extension. > * config/riscv/riscv-c.cc (riscv_cpu_cpp_builtins): > Add test marco. > * config/riscv/riscv.opt: Add new mask. > > gcc/testsuite/ChangeLog: > > * gcc.target/riscv/predef-__riscv_th_v_intrinsic.c: New test. > * gcc.target/riscv/rvv/xtheadvector.c: New test. > > Co-authored-by: Jin Ma <jinma@linux.alibaba.com> > Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> > Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> > --- > gcc/common/config/riscv/riscv-common.cc | 23 +++++++++++++++++++ > gcc/config/riscv/riscv-c.cc | 8 +++++-- > gcc/config/riscv/riscv.opt | 2 ++ > .../riscv/predef-__riscv_th_v_intrinsic.c | 11 +++++++++ > .../gcc.target/riscv/rvv/xtheadvector.c | 13 +++++++++++ > 5 files changed, 55 insertions(+), 2 deletions(-) > create mode 100644 gcc/testsuite/gcc.target/riscv/predef-__riscv_th_v_intrinsic.c > create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector.c > > diff --git a/gcc/common/config/riscv/riscv-common.cc b/gcc/common/config/riscv/riscv-common.cc > index 0301d170a41..449722070d4 100644 > --- a/gcc/common/config/riscv/riscv-common.cc > +++ b/gcc/common/config/riscv/riscv-common.cc > @@ -368,6 +368,7 @@ static const struct riscv_ext_version riscv_ext_version_table[] = > {"xtheadmemidx", ISA_SPEC_CLASS_NONE, 1, 0}, > {"xtheadmempair", ISA_SPEC_CLASS_NONE, 1, 0}, > {"xtheadsync", ISA_SPEC_CLASS_NONE, 1, 0}, > + {"xtheadvector", ISA_SPEC_CLASS_NONE, 1, 0}, > {"xventanacondops", ISA_SPEC_CLASS_NONE, 1, 0}, > @@ -1251,6 +1252,15 @@ riscv_subset_list::check_conflict_ext () > if (lookup ("zcmp")) > error_at (m_loc, "%<-march=%s%>: zcd conflicts with zcmp", m_arch); > } > + > + if ((lookup ("v") || lookup ("zve32x") > + || lookup ("zve64x") || lookup ("zve32f") > + || lookup ("zve64f") || lookup ("zve64d") > + || lookup ("zvl32b") || lookup ("zvl64b") > + || lookup ("zvl128b") || lookup ("zvfh")) > + && lookup ("xtheadvector")) > + error_at (m_loc, "%<-march=%s%>: xtheadvector conflicts with vector " > + "extension or its sub-extensions", m_arch); > } > /* Parsing function for multi-letter extensions. > @@ -1743,6 +1753,19 @@ static const riscv_ext_flag_table_t riscv_ext_flag_table[] = > {"xtheadmemidx", &gcc_options::x_riscv_xthead_subext, MASK_XTHEADMEMIDX}, > {"xtheadmempair", &gcc_options::x_riscv_xthead_subext, MASK_XTHEADMEMPAIR}, > {"xtheadsync", &gcc_options::x_riscv_xthead_subext, MASK_XTHEADSYNC}, > + {"xtheadvector", &gcc_options::x_riscv_xthead_subext, MASK_XTHEADVECTOR}, > + {"xtheadvector", &gcc_options::x_riscv_vector_elen_flags, MASK_VECTOR_ELEN_32}, > + {"xtheadvector", &gcc_options::x_riscv_vector_elen_flags, MASK_VECTOR_ELEN_64}, > + {"xtheadvector", &gcc_options::x_riscv_vector_elen_flags, MASK_VECTOR_ELEN_FP_32}, > + {"xtheadvector", &gcc_options::x_riscv_vector_elen_flags, MASK_VECTOR_ELEN_FP_64}, > + {"xtheadvector", &gcc_options::x_riscv_vector_elen_flags, MASK_VECTOR_ELEN_FP_16}, > + {"xtheadvector", &gcc_options::x_riscv_zvl_flags, MASK_ZVL32B}, > + {"xtheadvector", &gcc_options::x_riscv_zvl_flags, MASK_ZVL64B}, > + {"xtheadvector", &gcc_options::x_riscv_zvl_flags, MASK_ZVL128B}, > + {"xtheadvector", &gcc_options::x_riscv_zf_subext, MASK_ZVFHMIN}, > + {"xtheadvector", &gcc_options::x_riscv_zf_subext, MASK_ZVFH}, > + {"xtheadvector", &gcc_options::x_target_flags, MASK_FULL_V}, > + {"xtheadvector", &gcc_options::x_target_flags, MASK_VECTOR}, > {"xventanacondops", &gcc_options::x_riscv_xventana_subext, MASK_XVENTANACONDOPS}, > diff --git a/gcc/config/riscv/riscv-c.cc b/gcc/config/riscv/riscv-c.cc > index ba60cd8b555..422ddc2c308 100644 > --- a/gcc/config/riscv/riscv-c.cc > +++ b/gcc/config/riscv/riscv-c.cc > @@ -142,6 +142,10 @@ riscv_cpu_cpp_builtins (cpp_reader *pfile) > riscv_ext_version_value (0, 11)); > } > + if (TARGET_XTHEADVECTOR) > + builtin_define_with_int_value ("__riscv_th_v_intrinsic", > + riscv_ext_version_value (0, 11)); > + > /* Define architecture extension test macros. */ > builtin_define_with_int_value ("__riscv_arch_test", 1); > @@ -195,8 +199,8 @@ riscv_pragma_intrinsic (cpp_reader *) > { > if (!TARGET_VECTOR) > { > - error ("%<#pragma riscv intrinsic%> option %qs needs 'V' extension " > - "enabled", > + error ("%<#pragma riscv intrinsic%> option %qs needs 'V' or " > + "'XTHEADVECTOR' extension enabled", > name); > return; > } > diff --git a/gcc/config/riscv/riscv.opt b/gcc/config/riscv/riscv.opt > index 44ed6d69da2..bb18a22b693 100644 > --- a/gcc/config/riscv/riscv.opt > +++ b/gcc/config/riscv/riscv.opt > @@ -452,6 +452,8 @@ Mask(XTHEADMEMPAIR) Var(riscv_xthead_subext) > Mask(XTHEADSYNC) Var(riscv_xthead_subext) > +Mask(XTHEADVECTOR) Var(riscv_xthead_subext) > + > TargetVariable > int riscv_xventana_subext > diff --git a/gcc/testsuite/gcc.target/riscv/predef-__riscv_th_v_intrinsic.c b/gcc/testsuite/gcc.target/riscv/predef-__riscv_th_v_intrinsic.c > new file mode 100644 > index 00000000000..550b9039a06 > --- /dev/null > +++ b/gcc/testsuite/gcc.target/riscv/predef-__riscv_th_v_intrinsic.c > @@ -0,0 +1,11 @@ > +/* { dg-do compile } */ > +/* { dg-options "-march=rv64imafdcxtheadvector -mabi=lp64d" } */ > + > +int main () { > + > +#if __riscv_th_v_intrinsic != 11000 > +#error "__riscv_th_v_intrinsic" > +#endif > + > + return 0; > +} > diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector.c > new file mode 100644 > index 00000000000..8ad370172e3 > --- /dev/null > +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector.c > @@ -0,0 +1,13 @@ > +/* { dg-do compile } */ > +/* { dg-options "-march=rv32gc_xtheadvector" { target { rv32 } } } */ > +/* { dg-options "-march=rv64gc_xtheadvector" { target { rv64 } } } */ > + > +#ifndef __riscv_xtheadvector > +#error "Feature macro not defined" > +#endif > + > +int > +foo (int a) > +{ > + return a; > +} > -- > 2.17.1 > > ^ permalink raw reply [flat|nested] 133+ messages in thread
* [PATCH v5] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. 2024-01-12 3:18 ` [PATCH v5] RISC-V: Support XTheadVector extension Jun Sha (Joshua) 2024-01-12 3:20 ` [PATCH v4] RISC-V: Introduce XTheadVector as a subset of V1.0.0 Jun Sha (Joshua) @ 2024-01-12 3:21 ` Jun Sha (Joshua) 2024-01-12 7:31 ` juzhe.zhong 2024-01-12 3:22 ` [PATCH v6] RISC-V: Handle differences between XTheadvector and Vector Jun Sha (Joshua) ` (4 subsequent siblings) 6 siblings, 1 reply; 133+ messages in thread From: Jun Sha (Joshua) @ 2024-01-12 3:21 UTC (permalink / raw) To: gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, Jun Sha (Joshua), Jin Ma, Xianmiao Qu This patch adds th. prefix to all XTheadVector instructions by implementing new assembly output functions. We only check the prefix is 'v', so that no extra attribute is needed. gcc/ChangeLog: * config/riscv/riscv-protos.h (riscv_asm_output_opcode): Add new function to add assembler insn code prefix/suffix. (th_asm_output_opcode): Add Thead function to add assembler insn code prefix/suffix. * config/riscv/riscv.cc (riscv_asm_output_opcode): Implement function to add assembler insn code prefix/suffix. * config/riscv/riscv.h (ASM_OUTPUT_OPCODE): Add new function to add assembler insn code prefix/suffix. * config/riscv/thead.cc (th_asm_output_opcode): Implement Thead function to add assembler insn code prefix/suffix. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/prefix.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config/riscv/riscv-protos.h | 2 ++ gcc/config/riscv/riscv.cc | 11 +++++++++++ gcc/config/riscv/riscv.h | 4 ++++ gcc/config/riscv/thead.cc | 13 +++++++++++++ .../gcc.target/riscv/rvv/xtheadvector/prefix.c | 12 ++++++++++++ 5 files changed, 42 insertions(+) create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c diff --git a/gcc/config/riscv/riscv-protos.h b/gcc/config/riscv/riscv-protos.h index 31049ef7523..71724dabdb5 100644 --- a/gcc/config/riscv/riscv-protos.h +++ b/gcc/config/riscv/riscv-protos.h @@ -102,6 +102,7 @@ struct riscv_address_info { }; /* Routines implemented in riscv.cc. */ +extern const char *riscv_asm_output_opcode (FILE *asm_out_file, const char *p); extern enum riscv_symbol_type riscv_classify_symbolic_expression (rtx); extern bool riscv_symbolic_constant_p (rtx, enum riscv_symbol_type *); extern int riscv_float_const_rtx_index_for_fli (rtx); @@ -717,6 +718,7 @@ extern void th_mempair_prepare_save_restore_operands (rtx[4], bool, int, HOST_WIDE_INT, int, HOST_WIDE_INT); extern void th_mempair_save_restore_regs (rtx[4], bool, machine_mode); +extern const char *th_asm_output_opcode (FILE *asm_out_file, const char *p); #ifdef RTX_CODE extern const char* th_mempair_output_move (rtx[4], bool, machine_mode, RTX_CODE); diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index 0d1cbc5cb5f..51878797287 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -5636,6 +5636,17 @@ riscv_get_v_regno_alignment (machine_mode mode) return lmul; } +/* Define ASM_OUTPUT_OPCODE to do anything special before + emitting an opcode. */ +const char * +riscv_asm_output_opcode (FILE *asm_out_file, const char *p) +{ + if (TARGET_XTHEADVECTOR) + return th_asm_output_opcode (asm_out_file, p); + + return p; +} + /* Implement TARGET_PRINT_OPERAND. The RISCV-specific operand codes are: 'h' Print the high-part relocation associated with OP, after stripping diff --git a/gcc/config/riscv/riscv.h b/gcc/config/riscv/riscv.h index 6df9ec73c5e..c33361a254d 100644 --- a/gcc/config/riscv/riscv.h +++ b/gcc/config/riscv/riscv.h @@ -826,6 +826,10 @@ extern enum riscv_cc get_riscv_cc (const rtx use); asm_fprintf ((FILE), "%U%s", (NAME)); \ } while (0) +#undef ASM_OUTPUT_OPCODE +#define ASM_OUTPUT_OPCODE(STREAM, PTR) \ + (PTR) = riscv_asm_output_opcode(STREAM, PTR) + #define JUMP_TABLES_IN_TEXT_SECTION 0 #define CASE_VECTOR_MODE SImode #define CASE_VECTOR_PC_RELATIVE (riscv_cmodel != CM_MEDLOW) diff --git a/gcc/config/riscv/thead.cc b/gcc/config/riscv/thead.cc index 20353995931..dc3aed3904d 100644 --- a/gcc/config/riscv/thead.cc +++ b/gcc/config/riscv/thead.cc @@ -883,6 +883,19 @@ th_output_move (rtx dest, rtx src) return NULL; } +/* Define ASM_OUTPUT_OPCODE to do anything special before + emitting an opcode. */ +const char * +th_asm_output_opcode (FILE *asm_out_file, const char *p) +{ + /* We need to add th. prefix to all the xtheadvector + instructions here.*/ + if (current_output_insn != NULL && p[0] == 'v') + fputs ("th.", asm_out_file); + + return p; +} + /* Implement TARGET_PRINT_OPERAND_ADDRESS for XTheadMemIdx. */ bool diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c new file mode 100644 index 00000000000..eee727ef6b4 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c @@ -0,0 +1,12 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gc_xtheadvector -mabi=ilp32 -O0" } */ + +#include "riscv_vector.h" + +vint32m1_t +prefix (vint32m1_t vx, vint32m1_t vy, size_t vl) +{ + return __riscv_vadd_vv_i32m1 (vx, vy, vl); +} + +/* { dg-final { scan-assembler {\mth\.v\M} } } */ -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: [PATCH v5] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. 2024-01-12 3:21 ` [PATCH v5] RISC-V: Adds the prefix "th." for the instructions of XTheadVector Jun Sha (Joshua) @ 2024-01-12 7:31 ` juzhe.zhong 0 siblings, 0 replies; 133+ messages in thread From: juzhe.zhong @ 2024-01-12 7:31 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, cooper.joshua, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 5307 bytes --] OK. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-12 11:21 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Adds the prefix "th." for the instructions of XTheadVector. This patch adds th. prefix to all XTheadVector instructions by implementing new assembly output functions. We only check the prefix is 'v', so that no extra attribute is needed. gcc/ChangeLog: * config/riscv/riscv-protos.h (riscv_asm_output_opcode): Add new function to add assembler insn code prefix/suffix. (th_asm_output_opcode): Add Thead function to add assembler insn code prefix/suffix. * config/riscv/riscv.cc (riscv_asm_output_opcode): Implement function to add assembler insn code prefix/suffix. * config/riscv/riscv.h (ASM_OUTPUT_OPCODE): Add new function to add assembler insn code prefix/suffix. * config/riscv/thead.cc (th_asm_output_opcode): Implement Thead function to add assembler insn code prefix/suffix. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/prefix.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config/riscv/riscv-protos.h | 2 ++ gcc/config/riscv/riscv.cc | 11 +++++++++++ gcc/config/riscv/riscv.h | 4 ++++ gcc/config/riscv/thead.cc | 13 +++++++++++++ .../gcc.target/riscv/rvv/xtheadvector/prefix.c | 12 ++++++++++++ 5 files changed, 42 insertions(+) create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c diff --git a/gcc/config/riscv/riscv-protos.h b/gcc/config/riscv/riscv-protos.h index 31049ef7523..71724dabdb5 100644 --- a/gcc/config/riscv/riscv-protos.h +++ b/gcc/config/riscv/riscv-protos.h @@ -102,6 +102,7 @@ struct riscv_address_info { }; /* Routines implemented in riscv.cc. */ +extern const char *riscv_asm_output_opcode (FILE *asm_out_file, const char *p); extern enum riscv_symbol_type riscv_classify_symbolic_expression (rtx); extern bool riscv_symbolic_constant_p (rtx, enum riscv_symbol_type *); extern int riscv_float_const_rtx_index_for_fli (rtx); @@ -717,6 +718,7 @@ extern void th_mempair_prepare_save_restore_operands (rtx[4], bool, int, HOST_WIDE_INT, int, HOST_WIDE_INT); extern void th_mempair_save_restore_regs (rtx[4], bool, machine_mode); +extern const char *th_asm_output_opcode (FILE *asm_out_file, const char *p); #ifdef RTX_CODE extern const char* th_mempair_output_move (rtx[4], bool, machine_mode, RTX_CODE); diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index 0d1cbc5cb5f..51878797287 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -5636,6 +5636,17 @@ riscv_get_v_regno_alignment (machine_mode mode) return lmul; } +/* Define ASM_OUTPUT_OPCODE to do anything special before + emitting an opcode. */ +const char * +riscv_asm_output_opcode (FILE *asm_out_file, const char *p) +{ + if (TARGET_XTHEADVECTOR) + return th_asm_output_opcode (asm_out_file, p); + + return p; +} + /* Implement TARGET_PRINT_OPERAND. The RISCV-specific operand codes are: 'h' Print the high-part relocation associated with OP, after stripping diff --git a/gcc/config/riscv/riscv.h b/gcc/config/riscv/riscv.h index 6df9ec73c5e..c33361a254d 100644 --- a/gcc/config/riscv/riscv.h +++ b/gcc/config/riscv/riscv.h @@ -826,6 +826,10 @@ extern enum riscv_cc get_riscv_cc (const rtx use); asm_fprintf ((FILE), "%U%s", (NAME)); \ } while (0) +#undef ASM_OUTPUT_OPCODE +#define ASM_OUTPUT_OPCODE(STREAM, PTR) \ + (PTR) = riscv_asm_output_opcode(STREAM, PTR) + #define JUMP_TABLES_IN_TEXT_SECTION 0 #define CASE_VECTOR_MODE SImode #define CASE_VECTOR_PC_RELATIVE (riscv_cmodel != CM_MEDLOW) diff --git a/gcc/config/riscv/thead.cc b/gcc/config/riscv/thead.cc index 20353995931..dc3aed3904d 100644 --- a/gcc/config/riscv/thead.cc +++ b/gcc/config/riscv/thead.cc @@ -883,6 +883,19 @@ th_output_move (rtx dest, rtx src) return NULL; } +/* Define ASM_OUTPUT_OPCODE to do anything special before + emitting an opcode. */ +const char * +th_asm_output_opcode (FILE *asm_out_file, const char *p) +{ + /* We need to add th. prefix to all the xtheadvector + instructions here.*/ + if (current_output_insn != NULL && p[0] == 'v') + fputs ("th.", asm_out_file); + + return p; +} + /* Implement TARGET_PRINT_OPERAND_ADDRESS for XTheadMemIdx. */ bool diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c new file mode 100644 index 00000000000..eee727ef6b4 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/prefix.c @@ -0,0 +1,12 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gc_xtheadvector -mabi=ilp32 -O0" } */ + +#include "riscv_vector.h" + +vint32m1_t +prefix (vint32m1_t vx, vint32m1_t vy, size_t vl) +{ + return __riscv_vadd_vv_i32m1 (vx, vy, vl); +} + +/* { dg-final { scan-assembler {\mth\.v\M} } } */ -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* [PATCH v6] RISC-V: Handle differences between XTheadvector and Vector 2024-01-12 3:18 ` [PATCH v5] RISC-V: Support XTheadVector extension Jun Sha (Joshua) 2024-01-12 3:20 ` [PATCH v4] RISC-V: Introduce XTheadVector as a subset of V1.0.0 Jun Sha (Joshua) 2024-01-12 3:21 ` [PATCH v5] RISC-V: Adds the prefix "th." for the instructions of XTheadVector Jun Sha (Joshua) @ 2024-01-12 3:22 ` Jun Sha (Joshua) 2024-01-12 7:31 ` juzhe.zhong 2024-01-12 3:22 ` [PATCH v6] RISC-V: Add support for xtheadvector-specific intrinsics Jun Sha (Joshua) ` (3 subsequent siblings) 6 siblings, 1 reply; 133+ messages in thread From: Jun Sha (Joshua) @ 2024-01-12 3:22 UTC (permalink / raw) To: gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, Jun Sha (Joshua), Jin Ma, Xianmiao Qu This patch is to handle the differences in instruction generation between Vector and XTheadVector. In this version, we only support partial xtheadvector instructions that leverage directly from current RVV1.0 with simple adding "th." prefix. For different name xtheadvector instructions but share same patterns as RVV1.0 instructions, we will use ASM targethook to rewrite the whole string of the instructions in the following patches. For some vector patterns that cannot be avoided, we use "!TARGET_XTHEADVECTOR" to disable them in vector.md in order not to generate instructions that xtheadvector does not support, like vmv1r. gcc/ChangeLog: * config.gcc: Add files for XTheadVector intrinsics. * config/riscv/autovec.md: Guard XTheadVector. * config/riscv/predicates.md: Disable immediate vl for XTheadVector. * config/riscv/riscv-c.cc (riscv_pragma_intrinsic): Add pragma for XTheadVector. * config/riscv/riscv-string.cc (riscv_expand_block_move): Guard XTheadVector. * config/riscv/riscv-v.cc (vls_mode_valid_p): Avoid autovec. * config/riscv/riscv-vector-builtins-bases.cc: Do not normalize vsetvl instructions for XTheadVector. * config/riscv/riscv-vector-builtins-shapes.cc (check_type): (build_one): New check type function. * config/riscv/riscv-vector-switch.def (ENTRY): Disable fractional mode for the XTheadVector extension. (TUPLE_ENTRY): Likewise. * config/riscv/riscv.cc (riscv_v_adjust_bytesize): Guard XTheadVector. (riscv_preferred_simd_mode): Likewsie. (riscv_autovectorize_vector_modes): Likewise. (riscv_vector_mode_supported_any_target_p): Likewise. (TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P): Likewise. * config/riscv/thead.cc (th_asm_output_opcode): Rewrite vsetvl instructions. * config/riscv/vector.md: Include thead-vector.md and change fractional LMUL into 1 for vbool. * config/riscv/riscv_th_vector.h: New file. * config/riscv/thead-vector.md: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/base/pragma-1.c: Add XTheadVector. * gcc.target/riscv/rvv/base/abi-1.c: Exclude XTheadVector. * lib/target-supports.exp: Add target for XTheadVector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config.gcc | 2 +- gcc/config/riscv/autovec.md | 2 +- gcc/config/riscv/predicates.md | 4 +- gcc/config/riscv/riscv-c.cc | 3 +- gcc/config/riscv/riscv-string.cc | 3 +- gcc/config/riscv/riscv-v.cc | 2 +- .../riscv/riscv-vector-builtins-bases.cc | 48 ++++-- .../riscv/riscv-vector-builtins-shapes.cc | 23 +++ gcc/config/riscv/riscv-vector-switch.def | 150 +++++++++--------- gcc/config/riscv/riscv.cc | 20 ++- gcc/config/riscv/riscv_th_vector.h | 49 ++++++ gcc/config/riscv/thead-vector.md | 102 ++++++++++++ gcc/config/riscv/thead.cc | 23 ++- gcc/config/riscv/vector.md | 43 ++++- .../gcc.target/riscv/rvv/base/abi-1.c | 2 +- .../gcc.target/riscv/rvv/base/pragma-1.c | 2 +- gcc/testsuite/lib/target-supports.exp | 12 ++ 17 files changed, 380 insertions(+), 110 deletions(-) create mode 100644 gcc/config/riscv/riscv_th_vector.h create mode 100644 gcc/config/riscv/thead-vector.md diff --git a/gcc/config.gcc b/gcc/config.gcc index 7e583390024..047e4c02cf4 100644 --- a/gcc/config.gcc +++ b/gcc/config.gcc @@ -549,7 +549,7 @@ riscv*) extra_objs="${extra_objs} riscv-vector-builtins.o riscv-vector-builtins-shapes.o riscv-vector-builtins-bases.o" extra_objs="${extra_objs} thead.o riscv-target-attr.o" d_target_objs="riscv-d.o" - extra_headers="riscv_vector.h" + extra_headers="riscv_vector.h riscv_th_vector.h" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.cc" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.h" ;; diff --git a/gcc/config/riscv/autovec.md b/gcc/config/riscv/autovec.md index 775eaa825b0..0477781cabe 100644 --- a/gcc/config/riscv/autovec.md +++ b/gcc/config/riscv/autovec.md @@ -2579,7 +2579,7 @@ [(match_operand 0 "register_operand") (match_operand 1 "memory_operand") (match_operand:ANYI 2 "const_int_operand")] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" { riscv_vector::expand_rawmemchr(<MODE>mode, operands[0], operands[1], operands[2]); diff --git a/gcc/config/riscv/predicates.md b/gcc/config/riscv/predicates.md index b1a79cae50a..0337da88284 100644 --- a/gcc/config/riscv/predicates.md +++ b/gcc/config/riscv/predicates.md @@ -428,7 +428,9 @@ ;; Predicates for the V extension. (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (ior (match_test "TARGET_XTHEADVECTOR && rtx_equal_p (op, const0_rtx)") + (match_test "!TARGET_XTHEADVECTOR")) + (match_operand 0 "const_csr_operand")))) (define_special_predicate "autovec_length_operand" (ior (match_operand 0 "pmode_register_operand") diff --git a/gcc/config/riscv/riscv-c.cc b/gcc/config/riscv/riscv-c.cc index 422ddc2c308..2e306057347 100644 --- a/gcc/config/riscv/riscv-c.cc +++ b/gcc/config/riscv/riscv-c.cc @@ -195,7 +195,8 @@ riscv_pragma_intrinsic (cpp_reader *) const char *name = TREE_STRING_POINTER (x); - if (strcmp (name, "vector") == 0) + if (strcmp (name, "vector") == 0 + || strcmp (name, "xtheadvector") == 0) { if (!TARGET_VECTOR) { diff --git a/gcc/config/riscv/riscv-string.cc b/gcc/config/riscv/riscv-string.cc index f227b8667ce..b09b51d7526 100644 --- a/gcc/config/riscv/riscv-string.cc +++ b/gcc/config/riscv/riscv-string.cc @@ -773,7 +773,8 @@ riscv_expand_block_move_scalar (rtx dest, rtx src, rtx length) bool riscv_expand_block_move (rtx dest, rtx src, rtx length) { - if (TARGET_VECTOR && stringop_strategy & STRATEGY_VECTOR) + if ((TARGET_VECTOR && !TARGET_XTHEADVECTOR) + && stringop_strategy & STRATEGY_VECTOR) { bool ok = riscv_vector::expand_block_move (dest, src, length); if (ok) diff --git a/gcc/config/riscv/riscv-v.cc b/gcc/config/riscv/riscv-v.cc index 2491522191a..7e25deb966d 100644 --- a/gcc/config/riscv/riscv-v.cc +++ b/gcc/config/riscv/riscv-v.cc @@ -4392,7 +4392,7 @@ cmp_lmul_gt_one (machine_mode mode) bool vls_mode_valid_p (machine_mode vls_mode) { - if (!TARGET_VECTOR) + if (!TARGET_VECTOR || TARGET_XTHEADVECTOR) return false; if (riscv_autovec_preference == RVV_SCALABLE) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index fba98124473..1aa6e3c6665 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -131,23 +131,37 @@ public: tree type = builtin_types[e.type.index].vector; machine_mode mode = TYPE_MODE (type); - /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. - - - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 - - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 - - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 - - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 - - e8,m2/e16,m4/e32,m8 --> e8m2 - - e8,m4/e16,m8 --> e8m4 - - e8,m8 --> e8m8 - */ - /* SEW. */ - e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); - - /* LMUL. */ - machine_mode e8_mode - = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); - e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + + if (TARGET_XTHEADVECTOR) + { + machine_mode inner_mode = GET_MODE_INNER (mode); + /* SEW. */ + e.add_input_operand (Pmode, + gen_int_mode (GET_MODE_BITSIZE (inner_mode), Pmode)); + /* LMUL. */ + e.add_input_operand (Pmode, + gen_int_mode (get_vlmul (mode), Pmode)); + } + else + { + /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. + + - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 + - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 + - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 + - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 + - e8,m2/e16,m4/e32,m8 --> e8m2 + - e8,m4/e16,m8 --> e8m4 + - e8,m8 --> e8m8 + */ + /* SEW. */ + e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); + + /* LMUL. */ + machine_mode e8_mode + = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); + e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + } /* TAIL_ANY. */ e.add_input_operand (Pmode, diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index ee8058dcdb1..1e4f4d53de6 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -33,6 +33,25 @@ namespace riscv_vector { +/* Check whether the RETURN_TYPE and ARGUMENT_TYPES are + valid for the function. */ + +static bool +check_type (tree return_type, vec<tree> &argument_types) +{ + tree arg; + unsigned i; + + if (!return_type) + return false; + + FOR_EACH_VEC_ELT (argument_types, i, arg) + if (!arg) + return false; + + return true; +} + /* Add one function instance for GROUP, using operand suffix at index OI, mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ static void @@ -49,6 +68,10 @@ build_one (function_builder &b, const function_group_info &group, group.ops_infos.types[vec_type_idx].index); b.allocate_argument_types (function_instance, argument_types); b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + b.add_overloaded_function (function_instance, *group.shape); b.add_unique_function (function_instance, (*group.shape), return_type, argument_types); diff --git a/gcc/config/riscv/riscv-vector-switch.def b/gcc/config/riscv/riscv-vector-switch.def index 1ad26c2a3b2..452283b7416 100644 --- a/gcc/config/riscv/riscv-vector-switch.def +++ b/gcc/config/riscv/riscv-vector-switch.def @@ -68,9 +68,9 @@ Encode the ratio of SEW/LMUL into the mask types. #endif /* Disable modes if TARGET_MIN_VLEN == 32. */ -ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) -ENTRY (RVVMF32BI, true, LMUL_F4, 32) -ENTRY (RVVMF16BI, true, LMUL_F2, 16) +ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F8, 64) +ENTRY (RVVMF32BI, true, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F4, 32) +ENTRY (RVVMF16BI, true, TARGET_XTHEADVECTOR ? LMUL_1 : LMUL_F2 , 16) ENTRY (RVVMF8BI, true, LMUL_1, 8) ENTRY (RVVMF4BI, true, LMUL_2, 4) ENTRY (RVVMF2BI, true, LMUL_4, 2) @@ -81,39 +81,39 @@ ENTRY (RVVM8QI, true, LMUL_8, 1) ENTRY (RVVM4QI, true, LMUL_4, 2) ENTRY (RVVM2QI, true, LMUL_2, 4) ENTRY (RVVM1QI, true, LMUL_1, 8) -ENTRY (RVVMF2QI, true, LMUL_F2, 16) -ENTRY (RVVMF4QI, true, LMUL_F4, 32) -ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) +ENTRY (RVVMF2QI, !TARGET_XTHEADVECTOR, LMUL_F2, 16) +ENTRY (RVVMF4QI, !TARGET_XTHEADVECTOR, LMUL_F4, 32) +ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F8, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8HI, true, LMUL_8, 2) ENTRY (RVVM4HI, true, LMUL_4, 4) ENTRY (RVVM2HI, true, LMUL_2, 8) ENTRY (RVVM1HI, true, LMUL_1, 16) -ENTRY (RVVMF2HI, true, LMUL_F2, 32) -ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HI, !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_16. */ ENTRY (RVVM8HF, TARGET_VECTOR_ELEN_FP_16, LMUL_8, 2) ENTRY (RVVM4HF, TARGET_VECTOR_ELEN_FP_16, LMUL_4, 4) ENTRY (RVVM2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_2, 8) ENTRY (RVVM1HF, TARGET_VECTOR_ELEN_FP_16, LMUL_1, 16) -ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_F2, 32) -ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8SI, true, LMUL_8, 4) ENTRY (RVVM4SI, true, LMUL_4, 8) ENTRY (RVVM2SI, true, LMUL_2, 16) ENTRY (RVVM1SI, true, LMUL_1, 32) -ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_32. */ ENTRY (RVVM8SF, TARGET_VECTOR_ELEN_FP_32, LMUL_8, 4) ENTRY (RVVM4SF, TARGET_VECTOR_ELEN_FP_32, LMUL_4, 8) ENTRY (RVVM2SF, TARGET_VECTOR_ELEN_FP_32, LMUL_2, 16) ENTRY (RVVM1SF, TARGET_VECTOR_ELEN_FP_32, LMUL_1, 32) -ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if !TARGET_VECTOR_ELEN_64. */ ENTRY (RVVM8DI, TARGET_VECTOR_ELEN_64, LMUL_8, 8) @@ -140,127 +140,127 @@ ENTRY (RVVM1DF, TARGET_VECTOR_ELEN_FP_64, LMUL_1, 64) #endif TUPLE_ENTRY (RVVM1x8QI, true, RVVM1QI, 8, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x8QI, true, RVVMF2QI, 8, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x8QI, true, RVVMF4QI, 8, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 8, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x8QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 8, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x8QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 8, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 8, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x7QI, true, RVVM1QI, 7, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x7QI, true, RVVMF2QI, 7, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x7QI, true, RVVMF4QI, 7, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 7, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x7QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 7, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x7QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 7, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 7, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x6QI, true, RVVM1QI, 6, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x6QI, true, RVVMF2QI, 6, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x6QI, true, RVVMF4QI, 6, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 6, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x6QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 6, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x6QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 6, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 6, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x5QI, true, RVVM1QI, 5, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x5QI, true, RVVMF2QI, 5, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x5QI, true, RVVMF4QI, 5, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 5, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x5QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 5, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x5QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 5, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 5, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x4QI, true, RVVM2QI, 4, LMUL_2, 4) TUPLE_ENTRY (RVVM1x4QI, true, RVVM1QI, 4, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x4QI, true, RVVMF2QI, 4, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x4QI, true, RVVMF4QI, 4, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 4, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x4QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 4, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x4QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 4, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 4, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x3QI, true, RVVM2QI, 3, LMUL_2, 4) TUPLE_ENTRY (RVVM1x3QI, true, RVVM1QI, 3, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x3QI, true, RVVMF2QI, 3, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x3QI, true, RVVMF4QI, 3, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 3, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x3QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 3, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x3QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 3, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 3, LMUL_F8, 64) TUPLE_ENTRY (RVVM4x2QI, true, RVVM4QI, 2, LMUL_4, 2) TUPLE_ENTRY (RVVM2x2QI, true, RVVM2QI, 2, LMUL_2, 4) TUPLE_ENTRY (RVVM1x2QI, true, RVVM1QI, 2, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x2QI, true, RVVMF2QI, 2, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x2QI, true, RVVMF4QI, 2, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 2, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x2QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 2, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x2QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 2, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 2, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x8HI, true, RVVM1HI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HI, true, RVVMF2HI, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HI, true, RVVM1HI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HI, true, RVVMF2HI, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HI, true, RVVM1HI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HI, true, RVVMF2HI, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HI, true, RVVM1HI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HI, true, RVVMF2HI, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HI, true, RVVM2HI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HI, true, RVVM1HI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HI, true, RVVMF2HI, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HI, true, RVVM2HI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HI, true, RVVM1HI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HI, true, RVVMF2HI, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HI, true, RVVM4HI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HI, true, RVVM2HI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HI, true, RVVM1HI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HI, true, RVVMF2HI, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM4HF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8SI, true, RVVM1SI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SI, true, RVVM1SI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SI, true, RVVM1SI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SI, true, RVVM1SI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SI, true, RVVM2SI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SI, true, RVVM1SI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SI, true, RVVM2SI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SI, true, RVVM1SI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SI, true, RVVM4SI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SI, true, RVVM2SI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SI, true, RVVM1SI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM4SF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 8, LMUL_1, 16) TUPLE_ENTRY (RVVM1x7DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 7, LMUL_1, 16) diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index 8f74cd2e502..28cf1f414e9 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -1406,6 +1406,9 @@ riscv_v_adjust_bytesize (machine_mode mode, int scale) { if (riscv_v_ext_vector_mode_p (mode)) { + if (TARGET_XTHEADVECTOR) + return BYTES_PER_RISCV_VECTOR; + poly_int64 nunits = GET_MODE_NUNITS (mode); poly_int64 mode_size = GET_MODE_SIZE (mode); @@ -9995,7 +9998,7 @@ riscv_use_divmod_expander (void) static machine_mode riscv_preferred_simd_mode (scalar_mode mode) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::preferred_simd_mode (mode); return word_mode; @@ -10346,7 +10349,7 @@ riscv_mode_priority (int, int n) unsigned int riscv_autovectorize_vector_modes (vector_modes *modes, bool all) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::autovectorize_vector_modes (modes, all); return default_autovectorize_vector_modes (modes, all); @@ -10529,6 +10532,16 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) return false; } +/* Implements target hook vector_mode_supported_any_target_p. */ + +static bool +riscv_vector_mode_supported_any_target_p (machine_mode mode) +{ + if (TARGET_XTHEADVECTOR) + return false; + return true; +} + /* Initialize the GCC target structure. */ #undef TARGET_ASM_ALIGNED_HI_OP #define TARGET_ASM_ALIGNED_HI_OP "\t.half\t" @@ -10872,6 +10885,9 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) #undef TARGET_PREFERRED_ELSE_VALUE #define TARGET_PREFERRED_ELSE_VALUE riscv_preferred_else_value +#undef TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P +#define TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P riscv_vector_mode_supported_any_target_p + struct gcc_target targetm = TARGET_INITIALIZER; #include "gt-riscv.h" diff --git a/gcc/config/riscv/riscv_th_vector.h b/gcc/config/riscv/riscv_th_vector.h new file mode 100644 index 00000000000..b6b6738bdda --- /dev/null +++ b/gcc/config/riscv/riscv_th_vector.h @@ -0,0 +1,49 @@ +/* RISC-V 'XTheadVector' Extension intrinsics include file. + Copyright (C) 2024 Free Software Foundation, Inc. + + This file is part of GCC. + + GCC is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published + by the Free Software Foundation; either version 3, or (at your + option) any later version. + + GCC is distributed in the hope that it will be useful, but WITHOUT + ANY WARRANTY; without even the implied warranty of MERCHANTABILITY + or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public + License for more details. + + Under Section 7 of GPL version 3, you are granted additional + permissions described in the GCC Runtime Library Exception, version + 3.1, as published by the Free Software Foundation. + + You should have received a copy of the GNU General Public License and + a copy of the GCC Runtime Library Exception along with this program; + see the files COPYING3 and COPYING.RUNTIME respectively. If not, see + <http://www.gnu.org/licenses/>. */ + +#ifndef __RISCV_TH_VECTOR_H +#define __RISCV_TH_VECTOR_H + +#include <stdint.h> +#include <stddef.h> + +#ifndef __riscv_xtheadvector +#error "XTheadVector intrinsics require the xtheadvector extension." +#else +#ifdef __cplusplus +extern "C" { +#endif + +/* NOTE: This implementation of riscv_th_vector.h is intentionally short. It does + not define the RVV types and intrinsic functions directly in C and C++ + code, but instead uses the following pragma to tell GCC to insert the + necessary type and function definitions itself. The net effect is the + same, and the file is a complete implementation of riscv_th_vector.h. */ +#pragma riscv intrinsic "xtheadvector" + +#ifdef __cplusplus +} +#endif // __cplusplus +#endif // __riscv_xtheadvector +#endif // __RISCV_TH_ECTOR_H diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md new file mode 100644 index 00000000000..696b815252d --- /dev/null +++ b/gcc/config/riscv/thead-vector.md @@ -0,0 +1,102 @@ +(define_c_enum "unspec" [ + UNSPEC_TH_VWLDST +]) + +(define_mode_iterator V_VLS_VT [V VLS VT]) +(define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) + +(define_split + [(set (match_operand:V_VB_VLS_VT 0 "reg_or_mem_operand") + (match_operand:V_VB_VLS_VT 1 "reg_or_mem_operand"))] + "TARGET_XTHEADVECTOR" + [(const_int 0)] + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + }) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:V_VLS_VT 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:V_VLS_VT + [(match_operand:V_VLS_VT 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2")]) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:VB 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:VB + [(match_operand:VB 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2") + (set (attr "sew") (const_int 8)) + (set (attr "vlmul") (symbol_ref "riscv_vector::LMUL_1"))]) + +(define_insn_and_split "*pred_th_mov<mode>" + [(set (match_operand:VB_VLS 0 "nonimmediate_operand" "=vr, m, vr, vr, vr") + (if_then_else:VB_VLS + (unspec:VB_VLS + [(match_operand:VB_VLS 1 "vector_all_trues_mask_operand" "Wc1, Wc1, Wc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) + (match_operand:VB_VLS 3 "vector_move_operand" " m, vr, vr, Wc0, Wc1") + (match_operand:VB_VLS 2 "vector_undef_operand" " vu, vu, vu, vu, vu")))] + "TARGET_XTHEADVECTOR" + "@ + # + # + vmcpy.m\t%0,%3 + vmclr.m\t%0 + vmset.m\t%0" + "&& !reload_completed" + [(const_int 0)] + { + if ((MEM_P (operands[0]) || MEM_P (operands[3])) + || (REG_P (operands[0]) && REG_P (operands[3]) + && INTVAL (operands[5]) == riscv_vector::VLMAX)) + { + emit_move_insn (operands[0], operands[3]); + DONE; + } + FAIL; + } + [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/config/riscv/thead.cc b/gcc/config/riscv/thead.cc index 00b2e14c6c0..cc04101b256 100644 --- a/gcc/config/riscv/thead.cc +++ b/gcc/config/riscv/thead.cc @@ -27,6 +27,7 @@ #include "backend.h" #include "tree.h" #include "rtl.h" +#include "insn-attr.h" #include "explow.h" #include "memmodel.h" #include "emit-rtl.h" @@ -890,8 +891,26 @@ th_asm_output_opcode (FILE *asm_out_file, const char *p) { /* We need to add th. prefix to all the xtheadvector instructions here.*/ - if (current_output_insn != NULL && p[0] == 'v') - fputs ("th.", asm_out_file); + if (current_output_insn != NULL) + { + if (get_attr_type (current_output_insn) == TYPE_VSETVL) + { + if (strstr (p, "zero")) + { + if (strstr (p, "zero,zero")) + return "th.vsetvli\tzero,zero,e%0,%m1"; + else + return "th.vsetvli\tzero,%0,e%1,%m2"; + } + else + { + return "th.vsetvli\t%0,%1,e%2,%m3"; + } + } + + if (p[0] == 'v') + fputs ("th.", asm_out_file); + } return p; } diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index 24b7b4394be..34946090962 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -87,7 +87,7 @@ ;; check. However, we need default value of SEW for vsetvl instruction since there ;; is no field for ratio in the vsetvl instruction encoding. (define_attr "sew" "" - (cond [(eq_attr "mode" "RVVMF64BI,RVVMF32BI,RVVMF16BI,RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ + (cond [(eq_attr "mode" "RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ RVVM8QI,RVVM4QI,RVVM2QI,RVVM1QI,RVVMF2QI,RVVMF4QI,RVVMF8QI,\ RVVM1x8QI,RVVMF2x8QI,RVVMF4x8QI,RVVMF8x8QI,\ RVVM1x7QI,RVVMF2x7QI,RVVMF4x7QI,RVVMF8x7QI,\ @@ -99,6 +99,18 @@ V1QI,V2QI,V4QI,V8QI,V16QI,V32QI,V64QI,V128QI,V256QI,V512QI,V1024QI,V2048QI,V4096QI,\ V1BI,V2BI,V4BI,V8BI,V16BI,V32BI,V64BI,V128BI,V256BI,V512BI,V1024BI,V2048BI,V4096BI") (const_int 8) + (eq_attr "mode" "RVVMF16BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 16) + (const_int 8)) + (eq_attr "mode" "RVVMF32BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 32) + (const_int 8)) + (eq_attr "mode" "RVVMF64BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 64) + (const_int 8)) (eq_attr "mode" "RVVM8HI,RVVM4HI,RVVM2HI,RVVM1HI,RVVMF2HI,RVVMF4HI,\ RVVM1x8HI,RVVMF2x8HI,RVVMF4x8HI,\ RVVM1x7HI,RVVMF2x7HI,RVVMF4x7HI,\ @@ -159,9 +171,9 @@ (eq_attr "mode" "RVVM4QI,RVVMF2BI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2QI,RVVMF4BI") (symbol_ref "riscv_vector::LMUL_2") (eq_attr "mode" "RVVM1QI,RVVMF8BI") (symbol_ref "riscv_vector::LMUL_1") - (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "riscv_vector::LMUL_F2") - (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "riscv_vector::LMUL_F4") - (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "riscv_vector::LMUL_F8") + (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F2") + (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F4") + (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F8") (eq_attr "mode" "RVVM8HI") (symbol_ref "riscv_vector::LMUL_8") (eq_attr "mode" "RVVM4HI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2HI") (symbol_ref "riscv_vector::LMUL_2") @@ -436,6 +448,10 @@ vaeskf1,vaeskf2,vaesz,vsha2ms,vsha2ch,vsha2cl,vsm4k,vsm4r,\ vsm3me,vsm3c") (const_int INVALID_ATTRIBUTE) + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE) (eq_attr "mode" "RVVM8QI,RVVM1BI") (const_int 1) (eq_attr "mode" "RVVM4QI,RVVMF2BI") (const_int 2) (eq_attr "mode" "RVVM2QI,RVVMF4BI") (const_int 4) @@ -905,6 +921,8 @@ (symbol_ref "riscv_vector::FRM_DYN")] (symbol_ref "riscv_vector::FRM_NONE"))) +(include "thead-vector.md") + ;; ----------------------------------------------------------------- ;; ---- Miscellaneous Operations ;; ----------------------------------------------------------------- @@ -1078,6 +1096,12 @@ - We can not leave it to TARGET_SECONDARY_RELOAD since it happens before spilling. The clobber scratch is used by spilling fractional registers in IRA/LRA so it's too early. */ + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; @@ -1114,7 +1138,7 @@ (define_insn "*mov<mode>_whole" [(set (match_operand:V_WHOLE 0 "reg_or_mem_operand" "=vr, m,vr") (match_operand:V_WHOLE 1 "reg_or_mem_operand" " m,vr,vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "@ vl%m1re<sew>.v\t%0,%1 vs%m1r.v\t%1,%0 @@ -1135,6 +1159,13 @@ (match_operand:VB 1 "general_operand"))] "TARGET_VECTOR" { + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } + if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; }) @@ -1142,7 +1173,7 @@ (define_insn "*mov<mode>" [(set (match_operand:VB 0 "register_operand" "=vr") (match_operand:VB 1 "register_operand" " vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "vmv1r.v\t%0,%1" [(set_attr "type" "vmov") (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c index 2e0e12aa045..2eef9e1e1a8 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c @@ -1,4 +1,4 @@ -/* { dg-do compile } */ +/* { dg-do compile { target { ! riscv_xtheadvector } } } */ /* { dg-skip-if "test rvv intrinsic" { *-*-* } { "*" } { "-march=rv*v*" } } */ void foo0 () {__rvv_bool64_t t;} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c index 3d81b179235..ef329e30785 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c @@ -1,4 +1,4 @@ /* { dg-do compile } */ /* { dg-options "-O3 -march=rv32gc -mabi=ilp32d" } */ -#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' extension enabled} } */ +#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' or 'XTHEADVECTOR' extension enabled} } */ diff --git a/gcc/testsuite/lib/target-supports.exp b/gcc/testsuite/lib/target-supports.exp index b27c30b8c51..bf0ad5342d6 100644 --- a/gcc/testsuite/lib/target-supports.exp +++ b/gcc/testsuite/lib/target-supports.exp @@ -1956,6 +1956,18 @@ proc check_effective_target_riscv_zbb { } { }] } +# Return 1 if the target arch supports the XTheadVector extension, 0 otherwise. +# Cache the result. + +proc check_effective_target_riscv_xtheadvector { } { + return [check_no_compiler_messages riscv_ext_xtheadvector assembly { + #ifndef __riscv_xtheadvector + #error "Not __riscv_xtheadvector" + #endif + }] +} + + # Return 1 if we can execute code when using dg-add-options riscv_v proc check_effective_target_riscv_v_ok { } { -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: [PATCH v6] RISC-V: Handle differences between XTheadvector and Vector 2024-01-12 3:22 ` [PATCH v6] RISC-V: Handle differences between XTheadvector and Vector Jun Sha (Joshua) @ 2024-01-12 7:31 ` juzhe.zhong 0 siblings, 0 replies; 133+ messages in thread From: juzhe.zhong @ 2024-01-12 7:31 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, cooper.joshua, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 44081 bytes --] ok juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-12 11:22 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v6] RISC-V: Handle differences between XTheadvector and Vector This patch is to handle the differences in instruction generation between Vector and XTheadVector. In this version, we only support partial xtheadvector instructions that leverage directly from current RVV1.0 with simple adding "th." prefix. For different name xtheadvector instructions but share same patterns as RVV1.0 instructions, we will use ASM targethook to rewrite the whole string of the instructions in the following patches. For some vector patterns that cannot be avoided, we use "!TARGET_XTHEADVECTOR" to disable them in vector.md in order not to generate instructions that xtheadvector does not support, like vmv1r. gcc/ChangeLog: * config.gcc: Add files for XTheadVector intrinsics. * config/riscv/autovec.md: Guard XTheadVector. * config/riscv/predicates.md: Disable immediate vl for XTheadVector. * config/riscv/riscv-c.cc (riscv_pragma_intrinsic): Add pragma for XTheadVector. * config/riscv/riscv-string.cc (riscv_expand_block_move): Guard XTheadVector. * config/riscv/riscv-v.cc (vls_mode_valid_p): Avoid autovec. * config/riscv/riscv-vector-builtins-bases.cc: Do not normalize vsetvl instructions for XTheadVector. * config/riscv/riscv-vector-builtins-shapes.cc (check_type): (build_one): New check type function. * config/riscv/riscv-vector-switch.def (ENTRY): Disable fractional mode for the XTheadVector extension. (TUPLE_ENTRY): Likewise. * config/riscv/riscv.cc (riscv_v_adjust_bytesize): Guard XTheadVector. (riscv_preferred_simd_mode): Likewsie. (riscv_autovectorize_vector_modes): Likewise. (riscv_vector_mode_supported_any_target_p): Likewise. (TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P): Likewise. * config/riscv/thead.cc (th_asm_output_opcode): Rewrite vsetvl instructions. * config/riscv/vector.md: Include thead-vector.md and change fractional LMUL into 1 for vbool. * config/riscv/riscv_th_vector.h: New file. * config/riscv/thead-vector.md: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/base/pragma-1.c: Add XTheadVector. * gcc.target/riscv/rvv/base/abi-1.c: Exclude XTheadVector. * lib/target-supports.exp: Add target for XTheadVector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config.gcc | 2 +- gcc/config/riscv/autovec.md | 2 +- gcc/config/riscv/predicates.md | 4 +- gcc/config/riscv/riscv-c.cc | 3 +- gcc/config/riscv/riscv-string.cc | 3 +- gcc/config/riscv/riscv-v.cc | 2 +- .../riscv/riscv-vector-builtins-bases.cc | 48 ++++-- .../riscv/riscv-vector-builtins-shapes.cc | 23 +++ gcc/config/riscv/riscv-vector-switch.def | 150 +++++++++--------- gcc/config/riscv/riscv.cc | 20 ++- gcc/config/riscv/riscv_th_vector.h | 49 ++++++ gcc/config/riscv/thead-vector.md | 102 ++++++++++++ gcc/config/riscv/thead.cc | 23 ++- gcc/config/riscv/vector.md | 43 ++++- .../gcc.target/riscv/rvv/base/abi-1.c | 2 +- .../gcc.target/riscv/rvv/base/pragma-1.c | 2 +- gcc/testsuite/lib/target-supports.exp | 12 ++ 17 files changed, 380 insertions(+), 110 deletions(-) create mode 100644 gcc/config/riscv/riscv_th_vector.h create mode 100644 gcc/config/riscv/thead-vector.md diff --git a/gcc/config.gcc b/gcc/config.gcc index 7e583390024..047e4c02cf4 100644 --- a/gcc/config.gcc +++ b/gcc/config.gcc @@ -549,7 +549,7 @@ riscv*) extra_objs="${extra_objs} riscv-vector-builtins.o riscv-vector-builtins-shapes.o riscv-vector-builtins-bases.o" extra_objs="${extra_objs} thead.o riscv-target-attr.o" d_target_objs="riscv-d.o" - extra_headers="riscv_vector.h" + extra_headers="riscv_vector.h riscv_th_vector.h" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.cc" target_gtfiles="$target_gtfiles \$(srcdir)/config/riscv/riscv-vector-builtins.h" ;; diff --git a/gcc/config/riscv/autovec.md b/gcc/config/riscv/autovec.md index 775eaa825b0..0477781cabe 100644 --- a/gcc/config/riscv/autovec.md +++ b/gcc/config/riscv/autovec.md @@ -2579,7 +2579,7 @@ [(match_operand 0 "register_operand") (match_operand 1 "memory_operand") (match_operand:ANYI 2 "const_int_operand")] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" { riscv_vector::expand_rawmemchr(<MODE>mode, operands[0], operands[1], operands[2]); diff --git a/gcc/config/riscv/predicates.md b/gcc/config/riscv/predicates.md index b1a79cae50a..0337da88284 100644 --- a/gcc/config/riscv/predicates.md +++ b/gcc/config/riscv/predicates.md @@ -428,7 +428,9 @@ ;; Predicates for the V extension. (define_special_predicate "vector_length_operand" (ior (match_operand 0 "pmode_register_operand") - (match_operand 0 "const_csr_operand"))) + (and (ior (match_test "TARGET_XTHEADVECTOR && rtx_equal_p (op, const0_rtx)") + (match_test "!TARGET_XTHEADVECTOR")) + (match_operand 0 "const_csr_operand")))) (define_special_predicate "autovec_length_operand" (ior (match_operand 0 "pmode_register_operand") diff --git a/gcc/config/riscv/riscv-c.cc b/gcc/config/riscv/riscv-c.cc index 422ddc2c308..2e306057347 100644 --- a/gcc/config/riscv/riscv-c.cc +++ b/gcc/config/riscv/riscv-c.cc @@ -195,7 +195,8 @@ riscv_pragma_intrinsic (cpp_reader *) const char *name = TREE_STRING_POINTER (x); - if (strcmp (name, "vector") == 0) + if (strcmp (name, "vector") == 0 + || strcmp (name, "xtheadvector") == 0) { if (!TARGET_VECTOR) { diff --git a/gcc/config/riscv/riscv-string.cc b/gcc/config/riscv/riscv-string.cc index f227b8667ce..b09b51d7526 100644 --- a/gcc/config/riscv/riscv-string.cc +++ b/gcc/config/riscv/riscv-string.cc @@ -773,7 +773,8 @@ riscv_expand_block_move_scalar (rtx dest, rtx src, rtx length) bool riscv_expand_block_move (rtx dest, rtx src, rtx length) { - if (TARGET_VECTOR && stringop_strategy & STRATEGY_VECTOR) + if ((TARGET_VECTOR && !TARGET_XTHEADVECTOR) + && stringop_strategy & STRATEGY_VECTOR) { bool ok = riscv_vector::expand_block_move (dest, src, length); if (ok) diff --git a/gcc/config/riscv/riscv-v.cc b/gcc/config/riscv/riscv-v.cc index 2491522191a..7e25deb966d 100644 --- a/gcc/config/riscv/riscv-v.cc +++ b/gcc/config/riscv/riscv-v.cc @@ -4392,7 +4392,7 @@ cmp_lmul_gt_one (machine_mode mode) bool vls_mode_valid_p (machine_mode vls_mode) { - if (!TARGET_VECTOR) + if (!TARGET_VECTOR || TARGET_XTHEADVECTOR) return false; if (riscv_autovec_preference == RVV_SCALABLE) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index fba98124473..1aa6e3c6665 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -131,23 +131,37 @@ public: tree type = builtin_types[e.type.index].vector; machine_mode mode = TYPE_MODE (type); - /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. - - - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 - - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 - - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 - - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 - - e8,m2/e16,m4/e32,m8 --> e8m2 - - e8,m4/e16,m8 --> e8m4 - - e8,m8 --> e8m8 - */ - /* SEW. */ - e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); - - /* LMUL. */ - machine_mode e8_mode - = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); - e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + + if (TARGET_XTHEADVECTOR) + { + machine_mode inner_mode = GET_MODE_INNER (mode); + /* SEW. */ + e.add_input_operand (Pmode, + gen_int_mode (GET_MODE_BITSIZE (inner_mode), Pmode)); + /* LMUL. */ + e.add_input_operand (Pmode, + gen_int_mode (get_vlmul (mode), Pmode)); + } + else + { + /* Normalize same RATO (SEW/LMUL) into same vsetvl instruction. + + - e8,mf8/e16,mf4/e32,mf2/e64,m1 --> e8mf8 + - e8,mf4/e16,mf2/e32,m1/e64,m2 --> e8mf4 + - e8,mf2/e16,m1/e32,m2/e64,m4 --> e8mf2 + - e8,m1/e16,m2/e32,m4/e64,m8 --> e8m1 + - e8,m2/e16,m4/e32,m8 --> e8m2 + - e8,m4/e16,m8 --> e8m4 + - e8,m8 --> e8m8 + */ + /* SEW. */ + e.add_input_operand (Pmode, gen_int_mode (8, Pmode)); + + /* LMUL. */ + machine_mode e8_mode + = get_vector_mode (QImode, GET_MODE_NUNITS (mode)).require (); + e.add_input_operand (Pmode, gen_int_mode (get_vlmul (e8_mode), Pmode)); + } /* TAIL_ANY. */ e.add_input_operand (Pmode, diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index ee8058dcdb1..1e4f4d53de6 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -33,6 +33,25 @@ namespace riscv_vector { +/* Check whether the RETURN_TYPE and ARGUMENT_TYPES are + valid for the function. */ + +static bool +check_type (tree return_type, vec<tree> &argument_types) +{ + tree arg; + unsigned i; + + if (!return_type) + return false; + + FOR_EACH_VEC_ELT (argument_types, i, arg) + if (!arg) + return false; + + return true; +} + /* Add one function instance for GROUP, using operand suffix at index OI, mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ static void @@ -49,6 +68,10 @@ build_one (function_builder &b, const function_group_info &group, group.ops_infos.types[vec_type_idx].index); b.allocate_argument_types (function_instance, argument_types); b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + b.add_overloaded_function (function_instance, *group.shape); b.add_unique_function (function_instance, (*group.shape), return_type, argument_types); diff --git a/gcc/config/riscv/riscv-vector-switch.def b/gcc/config/riscv/riscv-vector-switch.def index 1ad26c2a3b2..452283b7416 100644 --- a/gcc/config/riscv/riscv-vector-switch.def +++ b/gcc/config/riscv/riscv-vector-switch.def @@ -68,9 +68,9 @@ Encode the ratio of SEW/LMUL into the mask types. #endif /* Disable modes if TARGET_MIN_VLEN == 32. */ -ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) -ENTRY (RVVMF32BI, true, LMUL_F4, 32) -ENTRY (RVVMF16BI, true, LMUL_F2, 16) +ENTRY (RVVMF64BI, TARGET_MIN_VLEN > 32, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F8, 64) +ENTRY (RVVMF32BI, true, TARGET_XTHEADVECTOR ? LMUL_1 :LMUL_F4, 32) +ENTRY (RVVMF16BI, true, TARGET_XTHEADVECTOR ? LMUL_1 : LMUL_F2 , 16) ENTRY (RVVMF8BI, true, LMUL_1, 8) ENTRY (RVVMF4BI, true, LMUL_2, 4) ENTRY (RVVMF2BI, true, LMUL_4, 2) @@ -81,39 +81,39 @@ ENTRY (RVVM8QI, true, LMUL_8, 1) ENTRY (RVVM4QI, true, LMUL_4, 2) ENTRY (RVVM2QI, true, LMUL_2, 4) ENTRY (RVVM1QI, true, LMUL_1, 8) -ENTRY (RVVMF2QI, true, LMUL_F2, 16) -ENTRY (RVVMF4QI, true, LMUL_F4, 32) -ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32, LMUL_F8, 64) +ENTRY (RVVMF2QI, !TARGET_XTHEADVECTOR, LMUL_F2, 16) +ENTRY (RVVMF4QI, !TARGET_XTHEADVECTOR, LMUL_F4, 32) +ENTRY (RVVMF8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F8, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8HI, true, LMUL_8, 2) ENTRY (RVVM4HI, true, LMUL_4, 4) ENTRY (RVVM2HI, true, LMUL_2, 8) ENTRY (RVVM1HI, true, LMUL_1, 16) -ENTRY (RVVMF2HI, true, LMUL_F2, 32) -ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HI, !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_16. */ ENTRY (RVVM8HF, TARGET_VECTOR_ELEN_FP_16, LMUL_8, 2) ENTRY (RVVM4HF, TARGET_VECTOR_ELEN_FP_16, LMUL_4, 4) ENTRY (RVVM2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_2, 8) ENTRY (RVVM1HF, TARGET_VECTOR_ELEN_FP_16, LMUL_1, 16) -ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16, LMUL_F2, 32) -ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, LMUL_F4, 64) +ENTRY (RVVMF2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, LMUL_F2, 32) +ENTRY (RVVMF4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F4, 64) /* Disable modes if TARGET_MIN_VLEN == 32. */ ENTRY (RVVM8SI, true, LMUL_8, 4) ENTRY (RVVM4SI, true, LMUL_4, 8) ENTRY (RVVM2SI, true, LMUL_2, 16) ENTRY (RVVM1SI, true, LMUL_1, 32) -ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if TARGET_MIN_VLEN == 32 or !TARGET_VECTOR_ELEN_FP_32. */ ENTRY (RVVM8SF, TARGET_VECTOR_ELEN_FP_32, LMUL_8, 4) ENTRY (RVVM4SF, TARGET_VECTOR_ELEN_FP_32, LMUL_4, 8) ENTRY (RVVM2SF, TARGET_VECTOR_ELEN_FP_32, LMUL_2, 16) ENTRY (RVVM1SF, TARGET_VECTOR_ELEN_FP_32, LMUL_1, 32) -ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, LMUL_F2, 64) +ENTRY (RVVMF2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, LMUL_F2, 64) /* Disable modes if !TARGET_VECTOR_ELEN_64. */ ENTRY (RVVM8DI, TARGET_VECTOR_ELEN_64, LMUL_8, 8) @@ -140,127 +140,127 @@ ENTRY (RVVM1DF, TARGET_VECTOR_ELEN_FP_64, LMUL_1, 64) #endif TUPLE_ENTRY (RVVM1x8QI, true, RVVM1QI, 8, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x8QI, true, RVVMF2QI, 8, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x8QI, true, RVVMF4QI, 8, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 8, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x8QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 8, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x8QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 8, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x8QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 8, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x7QI, true, RVVM1QI, 7, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x7QI, true, RVVMF2QI, 7, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x7QI, true, RVVMF4QI, 7, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 7, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x7QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 7, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x7QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 7, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x7QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 7, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x6QI, true, RVVM1QI, 6, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x6QI, true, RVVMF2QI, 6, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x6QI, true, RVVMF4QI, 6, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 6, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x6QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 6, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x6QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 6, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x6QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 6, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x5QI, true, RVVM1QI, 5, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x5QI, true, RVVMF2QI, 5, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x5QI, true, RVVMF4QI, 5, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 5, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x5QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 5, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x5QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 5, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x5QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 5, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x4QI, true, RVVM2QI, 4, LMUL_2, 4) TUPLE_ENTRY (RVVM1x4QI, true, RVVM1QI, 4, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x4QI, true, RVVMF2QI, 4, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x4QI, true, RVVMF4QI, 4, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 4, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x4QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 4, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x4QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 4, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x4QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 4, LMUL_F8, 64) TUPLE_ENTRY (RVVM2x3QI, true, RVVM2QI, 3, LMUL_2, 4) TUPLE_ENTRY (RVVM1x3QI, true, RVVM1QI, 3, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x3QI, true, RVVMF2QI, 3, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x3QI, true, RVVMF4QI, 3, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 3, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x3QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 3, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x3QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 3, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x3QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 3, LMUL_F8, 64) TUPLE_ENTRY (RVVM4x2QI, true, RVVM4QI, 2, LMUL_4, 2) TUPLE_ENTRY (RVVM2x2QI, true, RVVM2QI, 2, LMUL_2, 4) TUPLE_ENTRY (RVVM1x2QI, true, RVVM1QI, 2, LMUL_1, 8) -TUPLE_ENTRY (RVVMF2x2QI, true, RVVMF2QI, 2, LMUL_F2, 16) -TUPLE_ENTRY (RVVMF4x2QI, true, RVVMF4QI, 2, LMUL_F4, 32) -TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32, RVVMF8QI, 2, LMUL_F8, 64) +TUPLE_ENTRY (RVVMF2x2QI, !TARGET_XTHEADVECTOR, RVVMF2QI, 2, LMUL_F2, 16) +TUPLE_ENTRY (RVVMF4x2QI, !TARGET_XTHEADVECTOR, RVVMF4QI, 2, LMUL_F4, 32) +TUPLE_ENTRY (RVVMF8x2QI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF8QI, 2, LMUL_F8, 64) TUPLE_ENTRY (RVVM1x8HI, true, RVVM1HI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HI, true, RVVMF2HI, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HI, true, RVVM1HI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HI, true, RVVMF2HI, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HI, true, RVVM1HI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HI, true, RVVMF2HI, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HI, true, RVVM1HI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HI, true, RVVMF2HI, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HI, true, RVVM2HI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HI, true, RVVM1HI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HI, true, RVVMF2HI, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HI, true, RVVM2HI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HI, true, RVVM1HI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HI, true, RVVMF2HI, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HI, true, RVVM4HI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HI, true, RVVM2HI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HI, true, RVVM1HI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HI, true, RVVMF2HI, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32, RVVMF4HI, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HI, !TARGET_XTHEADVECTOR, RVVMF2HI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HI, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 8, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 8, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x8HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x8HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 8, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x7HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 7, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 7, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x7HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x7HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 7, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x6HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 6, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 6, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x6HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x6HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 6, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x5HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 5, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 5, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x5HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x5HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 5, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 4, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 4, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x4HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x4HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 4, LMUL_F4, 64) TUPLE_ENTRY (RVVM2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 3, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 3, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x3HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x3HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 3, LMUL_F4, 64) TUPLE_ENTRY (RVVM4x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM4HF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM2HF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2HF, TARGET_VECTOR_ELEN_FP_16, RVVM1HF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16, RVVMF2HF, 2, LMUL_F2, 32) -TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32, RVVMF4HF, 2, LMUL_F4, 64) +TUPLE_ENTRY (RVVMF2x2HF, TARGET_VECTOR_ELEN_FP_16 && !TARGET_XTHEADVECTOR, RVVMF2HF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF4x2HF, TARGET_VECTOR_ELEN_FP_16 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF4HF, 2, LMUL_F4, 64) TUPLE_ENTRY (RVVM1x8SI, true, RVVM1SI, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SI, true, RVVM1SI, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SI, (TARGET_MIN_VLEN > 32) && !TARGET_XTHEADVECTOR, RVVMF2SI, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SI, true, RVVM1SI, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SI, true, RVVM1SI, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SI, true, RVVM2SI, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SI, true, RVVM1SI, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SI, true, RVVM2SI, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SI, true, RVVM1SI, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SI, true, RVVM4SI, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SI, true, RVVM2SI, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SI, true, RVVM1SI, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32, RVVMF2SI, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SI, TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SI, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 8, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 8, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x8SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 8, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x7SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 7, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 7, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x7SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 7, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x6SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 6, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 6, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x6SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 6, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x5SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 5, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 5, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x5SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 5, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 4, LMUL_2, 8) TUPLE_ENTRY (RVVM1x4SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 4, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 4, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x4SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 4, LMUL_F2, 32) TUPLE_ENTRY (RVVM2x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 3, LMUL_2, 8) TUPLE_ENTRY (RVVM1x3SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 3, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 3, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x3SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 3, LMUL_F2, 32) TUPLE_ENTRY (RVVM4x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM4SF, 2, LMUL_4, 4) TUPLE_ENTRY (RVVM2x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM2SF, 2, LMUL_2, 8) TUPLE_ENTRY (RVVM1x2SF, TARGET_VECTOR_ELEN_FP_32, RVVM1SF, 2, LMUL_1, 16) -TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32, RVVMF2SF, 2, LMUL_F2, 32) +TUPLE_ENTRY (RVVMF2x2SF, TARGET_VECTOR_ELEN_FP_32 && TARGET_MIN_VLEN > 32 && !TARGET_XTHEADVECTOR, RVVMF2SF, 2, LMUL_F2, 32) TUPLE_ENTRY (RVVM1x8DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 8, LMUL_1, 16) TUPLE_ENTRY (RVVM1x7DI, TARGET_VECTOR_ELEN_64, RVVM1DI, 7, LMUL_1, 16) diff --git a/gcc/config/riscv/riscv.cc b/gcc/config/riscv/riscv.cc index 8f74cd2e502..28cf1f414e9 100644 --- a/gcc/config/riscv/riscv.cc +++ b/gcc/config/riscv/riscv.cc @@ -1406,6 +1406,9 @@ riscv_v_adjust_bytesize (machine_mode mode, int scale) { if (riscv_v_ext_vector_mode_p (mode)) { + if (TARGET_XTHEADVECTOR) + return BYTES_PER_RISCV_VECTOR; + poly_int64 nunits = GET_MODE_NUNITS (mode); poly_int64 mode_size = GET_MODE_SIZE (mode); @@ -9995,7 +9998,7 @@ riscv_use_divmod_expander (void) static machine_mode riscv_preferred_simd_mode (scalar_mode mode) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::preferred_simd_mode (mode); return word_mode; @@ -10346,7 +10349,7 @@ riscv_mode_priority (int, int n) unsigned int riscv_autovectorize_vector_modes (vector_modes *modes, bool all) { - if (TARGET_VECTOR) + if (TARGET_VECTOR && !TARGET_XTHEADVECTOR) return riscv_vector::autovectorize_vector_modes (modes, all); return default_autovectorize_vector_modes (modes, all); @@ -10529,6 +10532,16 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) return false; } +/* Implements target hook vector_mode_supported_any_target_p. */ + +static bool +riscv_vector_mode_supported_any_target_p (machine_mode mode) +{ + if (TARGET_XTHEADVECTOR) + return false; + return true; +} + /* Initialize the GCC target structure. */ #undef TARGET_ASM_ALIGNED_HI_OP #define TARGET_ASM_ALIGNED_HI_OP "\t.half\t" @@ -10872,6 +10885,9 @@ extract_base_offset_in_addr (rtx mem, rtx *base, rtx *offset) #undef TARGET_PREFERRED_ELSE_VALUE #define TARGET_PREFERRED_ELSE_VALUE riscv_preferred_else_value +#undef TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P +#define TARGET_VECTOR_MODE_SUPPORTED_ANY_TARGET_P riscv_vector_mode_supported_any_target_p + struct gcc_target targetm = TARGET_INITIALIZER; #include "gt-riscv.h" diff --git a/gcc/config/riscv/riscv_th_vector.h b/gcc/config/riscv/riscv_th_vector.h new file mode 100644 index 00000000000..b6b6738bdda --- /dev/null +++ b/gcc/config/riscv/riscv_th_vector.h @@ -0,0 +1,49 @@ +/* RISC-V 'XTheadVector' Extension intrinsics include file. + Copyright (C) 2024 Free Software Foundation, Inc. + + This file is part of GCC. + + GCC is free software; you can redistribute it and/or modify it + under the terms of the GNU General Public License as published + by the Free Software Foundation; either version 3, or (at your + option) any later version. + + GCC is distributed in the hope that it will be useful, but WITHOUT + ANY WARRANTY; without even the implied warranty of MERCHANTABILITY + or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public + License for more details. + + Under Section 7 of GPL version 3, you are granted additional + permissions described in the GCC Runtime Library Exception, version + 3.1, as published by the Free Software Foundation. + + You should have received a copy of the GNU General Public License and + a copy of the GCC Runtime Library Exception along with this program; + see the files COPYING3 and COPYING.RUNTIME respectively. If not, see + <http://www.gnu.org/licenses/>. */ + +#ifndef __RISCV_TH_VECTOR_H +#define __RISCV_TH_VECTOR_H + +#include <stdint.h> +#include <stddef.h> + +#ifndef __riscv_xtheadvector +#error "XTheadVector intrinsics require the xtheadvector extension." +#else +#ifdef __cplusplus +extern "C" { +#endif + +/* NOTE: This implementation of riscv_th_vector.h is intentionally short. It does + not define the RVV types and intrinsic functions directly in C and C++ + code, but instead uses the following pragma to tell GCC to insert the + necessary type and function definitions itself. The net effect is the + same, and the file is a complete implementation of riscv_th_vector.h. */ +#pragma riscv intrinsic "xtheadvector" + +#ifdef __cplusplus +} +#endif // __cplusplus +#endif // __riscv_xtheadvector +#endif // __RISCV_TH_ECTOR_H diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md new file mode 100644 index 00000000000..696b815252d --- /dev/null +++ b/gcc/config/riscv/thead-vector.md @@ -0,0 +1,102 @@ +(define_c_enum "unspec" [ + UNSPEC_TH_VWLDST +]) + +(define_mode_iterator V_VLS_VT [V VLS VT]) +(define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) + +(define_split + [(set (match_operand:V_VB_VLS_VT 0 "reg_or_mem_operand") + (match_operand:V_VB_VLS_VT 1 "reg_or_mem_operand"))] + "TARGET_XTHEADVECTOR" + [(const_int 0)] + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + }) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:V_VLS_VT 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:V_VLS_VT + [(match_operand:V_VLS_VT 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2")]) + +(define_insn_and_split "@pred_th_whole_mov<mode>" + [(set (match_operand:VB 0 "reg_or_mem_operand" "=vr,vr, m") + (unspec:VB + [(match_operand:VB 1 "reg_or_mem_operand" " vr, m,vr") + (match_operand 2 "vector_length_operand" " rK, rK, rK") + (match_operand 3 "const_1_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] + UNSPEC_TH_VWLDST))] + "TARGET_XTHEADVECTOR" + "@ + vmv.v.v\t%0,%1 + vle.v\t%0,%1 + vse.v\t%1,%0" + "&& REG_P (operands[0]) && REG_P (operands[1]) + && REGNO (operands[0]) == REGNO (operands[1])" + [(const_int 0)] + "" + [(set_attr "type" "vimov,vlds,vlds") + (set_attr "mode" "<MODE>") + (set (attr "ta") (symbol_ref "riscv_vector::TAIL_UNDISTURBED")) + (set (attr "ma") (symbol_ref "riscv_vector::MASK_UNDISTURBED")) + (set (attr "avl_type_idx") (const_int 3)) + (set_attr "vl_op_idx" "2") + (set (attr "sew") (const_int 8)) + (set (attr "vlmul") (symbol_ref "riscv_vector::LMUL_1"))]) + +(define_insn_and_split "*pred_th_mov<mode>" + [(set (match_operand:VB_VLS 0 "nonimmediate_operand" "=vr, m, vr, vr, vr") + (if_then_else:VB_VLS + (unspec:VB_VLS + [(match_operand:VB_VLS 1 "vector_all_trues_mask_operand" "Wc1, Wc1, Wc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) + (match_operand:VB_VLS 3 "vector_move_operand" " m, vr, vr, Wc0, Wc1") + (match_operand:VB_VLS 2 "vector_undef_operand" " vu, vu, vu, vu, vu")))] + "TARGET_XTHEADVECTOR" + "@ + # + # + vmcpy.m\t%0,%3 + vmclr.m\t%0 + vmset.m\t%0" + "&& !reload_completed" + [(const_int 0)] + { + if ((MEM_P (operands[0]) || MEM_P (operands[3])) + || (REG_P (operands[0]) && REG_P (operands[3]) + && INTVAL (operands[5]) == riscv_vector::VLMAX)) + { + emit_move_insn (operands[0], operands[3]); + DONE; + } + FAIL; + } + [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/config/riscv/thead.cc b/gcc/config/riscv/thead.cc index 00b2e14c6c0..cc04101b256 100644 --- a/gcc/config/riscv/thead.cc +++ b/gcc/config/riscv/thead.cc @@ -27,6 +27,7 @@ #include "backend.h" #include "tree.h" #include "rtl.h" +#include "insn-attr.h" #include "explow.h" #include "memmodel.h" #include "emit-rtl.h" @@ -890,8 +891,26 @@ th_asm_output_opcode (FILE *asm_out_file, const char *p) { /* We need to add th. prefix to all the xtheadvector instructions here.*/ - if (current_output_insn != NULL && p[0] == 'v') - fputs ("th.", asm_out_file); + if (current_output_insn != NULL) + { + if (get_attr_type (current_output_insn) == TYPE_VSETVL) + { + if (strstr (p, "zero")) + { + if (strstr (p, "zero,zero")) + return "th.vsetvli\tzero,zero,e%0,%m1"; + else + return "th.vsetvli\tzero,%0,e%1,%m2"; + } + else + { + return "th.vsetvli\t%0,%1,e%2,%m3"; + } + } + + if (p[0] == 'v') + fputs ("th.", asm_out_file); + } return p; } diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index 24b7b4394be..34946090962 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -87,7 +87,7 @@ ;; check. However, we need default value of SEW for vsetvl instruction since there ;; is no field for ratio in the vsetvl instruction encoding. (define_attr "sew" "" - (cond [(eq_attr "mode" "RVVMF64BI,RVVMF32BI,RVVMF16BI,RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ + (cond [(eq_attr "mode" "RVVMF8BI,RVVMF4BI,RVVMF2BI,RVVM1BI,\ RVVM8QI,RVVM4QI,RVVM2QI,RVVM1QI,RVVMF2QI,RVVMF4QI,RVVMF8QI,\ RVVM1x8QI,RVVMF2x8QI,RVVMF4x8QI,RVVMF8x8QI,\ RVVM1x7QI,RVVMF2x7QI,RVVMF4x7QI,RVVMF8x7QI,\ @@ -99,6 +99,18 @@ V1QI,V2QI,V4QI,V8QI,V16QI,V32QI,V64QI,V128QI,V256QI,V512QI,V1024QI,V2048QI,V4096QI,\ V1BI,V2BI,V4BI,V8BI,V16BI,V32BI,V64BI,V128BI,V256BI,V512BI,V1024BI,V2048BI,V4096BI") (const_int 8) + (eq_attr "mode" "RVVMF16BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 16) + (const_int 8)) + (eq_attr "mode" "RVVMF32BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 32) + (const_int 8)) + (eq_attr "mode" "RVVMF64BI") + (if_then_else (match_test "TARGET_XTHEADVECTOR") + (const_int 64) + (const_int 8)) (eq_attr "mode" "RVVM8HI,RVVM4HI,RVVM2HI,RVVM1HI,RVVMF2HI,RVVMF4HI,\ RVVM1x8HI,RVVMF2x8HI,RVVMF4x8HI,\ RVVM1x7HI,RVVMF2x7HI,RVVMF4x7HI,\ @@ -159,9 +171,9 @@ (eq_attr "mode" "RVVM4QI,RVVMF2BI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2QI,RVVMF4BI") (symbol_ref "riscv_vector::LMUL_2") (eq_attr "mode" "RVVM1QI,RVVMF8BI") (symbol_ref "riscv_vector::LMUL_1") - (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "riscv_vector::LMUL_F2") - (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "riscv_vector::LMUL_F4") - (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "riscv_vector::LMUL_F8") + (eq_attr "mode" "RVVMF2QI,RVVMF16BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F2") + (eq_attr "mode" "RVVMF4QI,RVVMF32BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F4") + (eq_attr "mode" "RVVMF8QI,RVVMF64BI") (symbol_ref "TARGET_XTHEADVECTOR ? riscv_vector::LMUL_1 : riscv_vector::LMUL_F8") (eq_attr "mode" "RVVM8HI") (symbol_ref "riscv_vector::LMUL_8") (eq_attr "mode" "RVVM4HI") (symbol_ref "riscv_vector::LMUL_4") (eq_attr "mode" "RVVM2HI") (symbol_ref "riscv_vector::LMUL_2") @@ -436,6 +448,10 @@ vaeskf1,vaeskf2,vaesz,vsha2ms,vsha2ch,vsha2cl,vsm4k,vsm4r,\ vsm3me,vsm3c") (const_int INVALID_ATTRIBUTE) + (and (eq_attr "type" "vlde,vste,vlsegde,vssegte,vlsegds,vssegts,\ + vlsegdff,vssegtux,vlsegdox,vlsegdux") + (match_test "TARGET_XTHEADVECTOR")) + (const_int INVALID_ATTRIBUTE) (eq_attr "mode" "RVVM8QI,RVVM1BI") (const_int 1) (eq_attr "mode" "RVVM4QI,RVVMF2BI") (const_int 2) (eq_attr "mode" "RVVM2QI,RVVMF4BI") (const_int 4) @@ -905,6 +921,8 @@ (symbol_ref "riscv_vector::FRM_DYN")] (symbol_ref "riscv_vector::FRM_NONE"))) +(include "thead-vector.md") + ;; ----------------------------------------------------------------- ;; ---- Miscellaneous Operations ;; ----------------------------------------------------------------- @@ -1078,6 +1096,12 @@ - We can not leave it to TARGET_SECONDARY_RELOAD since it happens before spilling. The clobber scratch is used by spilling fractional registers in IRA/LRA so it's too early. */ + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; @@ -1114,7 +1138,7 @@ (define_insn "*mov<mode>_whole" [(set (match_operand:V_WHOLE 0 "reg_or_mem_operand" "=vr, m,vr") (match_operand:V_WHOLE 1 "reg_or_mem_operand" " m,vr,vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "@ vl%m1re<sew>.v\t%0,%1 vs%m1r.v\t%1,%0 @@ -1135,6 +1159,13 @@ (match_operand:VB 1 "general_operand"))] "TARGET_VECTOR" { + if (TARGET_XTHEADVECTOR) + { + emit_insn (gen_pred_th_whole_mov (<MODE>mode, operands[0], operands[1], + RVV_VLMAX, GEN_INT(riscv_vector::VLMAX))); + DONE; + } + if (riscv_vector::legitimize_move (operands[0], &operands[1])) DONE; }) @@ -1142,7 +1173,7 @@ (define_insn "*mov<mode>" [(set (match_operand:VB 0 "register_operand" "=vr") (match_operand:VB 1 "register_operand" " vr"))] - "TARGET_VECTOR" + "TARGET_VECTOR && !TARGET_XTHEADVECTOR" "vmv1r.v\t%0,%1" [(set_attr "type" "vmov") (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c index 2e0e12aa045..2eef9e1e1a8 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/abi-1.c @@ -1,4 +1,4 @@ -/* { dg-do compile } */ +/* { dg-do compile { target { ! riscv_xtheadvector } } } */ /* { dg-skip-if "test rvv intrinsic" { *-*-* } { "*" } { "-march=rv*v*" } } */ void foo0 () {__rvv_bool64_t t;} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c index 3d81b179235..ef329e30785 100644 --- a/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c +++ b/gcc/testsuite/gcc.target/riscv/rvv/base/pragma-1.c @@ -1,4 +1,4 @@ /* { dg-do compile } */ /* { dg-options "-O3 -march=rv32gc -mabi=ilp32d" } */ -#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' extension enabled} } */ +#pragma riscv intrinsic "vector" /* { dg-error {#pragma riscv intrinsic' option 'vector' needs 'V' or 'XTHEADVECTOR' extension enabled} } */ diff --git a/gcc/testsuite/lib/target-supports.exp b/gcc/testsuite/lib/target-supports.exp index b27c30b8c51..bf0ad5342d6 100644 --- a/gcc/testsuite/lib/target-supports.exp +++ b/gcc/testsuite/lib/target-supports.exp @@ -1956,6 +1956,18 @@ proc check_effective_target_riscv_zbb { } { }] } +# Return 1 if the target arch supports the XTheadVector extension, 0 otherwise. +# Cache the result. + +proc check_effective_target_riscv_xtheadvector { } { + return [check_no_compiler_messages riscv_ext_xtheadvector assembly { + #ifndef __riscv_xtheadvector + #error "Not __riscv_xtheadvector" + #endif + }] +} + + # Return 1 if we can execute code when using dg-add-options riscv_v proc check_effective_target_riscv_v_ok { } { -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* [PATCH v6] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-12 3:18 ` [PATCH v5] RISC-V: Support XTheadVector extension Jun Sha (Joshua) ` (2 preceding siblings ...) 2024-01-12 3:22 ` [PATCH v6] RISC-V: Handle differences between XTheadvector and Vector Jun Sha (Joshua) @ 2024-01-12 3:22 ` Jun Sha (Joshua) 2024-01-12 7:32 ` juzhe.zhong 2024-01-12 3:23 ` [PATCH v6] RISC-V: Fix register overlap issue for some xtheadvector instructions Jun Sha (Joshua) ` (2 subsequent siblings) 6 siblings, 1 reply; 133+ messages in thread From: Jun Sha (Joshua) @ 2024-01-12 3:22 UTC (permalink / raw) To: gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, Jun Sha (Joshua), Jin Ma, Xianmiao Qu This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (class th_extract): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (struct th_extract_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): Redefine DEF_RVV_FUNCTION for XTheadVector special intrinsics. * config/riscv/riscv-vector-builtins.h (enum required_ext): Add new XTheadVector member. (struct function_group_info): Likewise. * config/riscv/t-riscv: Add thead-vector-builtins-functions.def * config/riscv/thead-vector.md (@pred_mov_width<vlmem_op_attr><mode>): Add new patterns. (*pred_mov_width<vlmem_op_attr><mode>): Likewise. (@pred_store_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_load_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_store_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_load_width<vlmem_op_attr><mode>): Likewise. (@pred_th_extract<mode>): Likewise. (*pred_th_extract<mode>): Likewise. * config/riscv/thead-vector-builtins-functions.def: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 +++ .../riscv/riscv-vector-builtins-shapes.cc | 160 +++++++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + gcc/config/riscv/riscv-vector-builtins.cc | 70 +++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 250 ++++++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 +++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 +++++ 15 files changed, 1104 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 1aa6e3c6665..b6f6e4ff37e 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2141,6 +2141,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2603,6 +2680,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> vsuxw_obj; +static CONSTEXPR const th_extract vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2894,6 +3002,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (vlb) +BASE (vlh) +BASE (vlw) +BASE (vlbu) +BASE (vlhu) +BASE (vlwu) +BASE (vsb) +BASE (vsh) +BASE (vsw) +BASE (vlsb) +BASE (vlsh) +BASE (vlsw) +BASE (vlsbu) +BASE (vlshu) +BASE (vlswu) +BASE (vssb) +BASE (vssh) +BASE (vssw) +BASE (vlxb) +BASE (vlxh) +BASE (vlxw) +BASE (vlxbu) +BASE (vlxhu) +BASE (vlxwu) +BASE (vsxb) +BASE (vsxh) +BASE (vsxw) +BASE (vsuxb) +BASE (vsuxh) +BASE (vsuxw) +BASE (vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 87c7f43a423..1f2c94d3541 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -280,6 +280,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const vlb; +extern const function_base *const vlh; +extern const function_base *const vlw; +extern const function_base *const vlbu; +extern const function_base *const vlhu; +extern const function_base *const vlwu; +extern const function_base *const vsb; +extern const function_base *const vsh; +extern const function_base *const vsw; +extern const function_base *const vlsb; +extern const function_base *const vlsh; +extern const function_base *const vlsw; +extern const function_base *const vlsbu; +extern const function_base *const vlshu; +extern const function_base *const vlswu; +extern const function_base *const vssb; +extern const function_base *const vssh; +extern const function_base *const vssw; +extern const function_base *const vlxb; +extern const function_base *const vlxh; +extern const function_base *const vlxw; +extern const function_base *const vlxbu; +extern const function_base *const vlxhu; +extern const function_base *const vlxwu; +extern const function_base *const vsxb; +extern const function_base *const vsxh; +extern const function_base *const vsxw; +extern const function_base *const vsuxb; +extern const function_base *const vsuxh; +extern const function_base *const vsuxw; +extern const function_base *const vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..8e90b17a94b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,146 @@ struct indexed_loadstore_def : public function_shape } }; +/* Add one function instance for GROUP, using operand suffix at index OI, + mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ +static void +build_th_loadstore (function_builder &b, const function_group_info &group, + unsigned int pred_idx, unsigned int vec_type_idx) +{ + auto_vec<tree, 5> argument_types; + function_instance function_instance (group.base_name, *group.base, + *group.shape, + group.ops_infos.types[vec_type_idx], + group.preds[pred_idx], &group.ops_infos); + tree return_type = group.ops_infos.ret.get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + b.allocate_argument_types (function_instance, argument_types); + b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + + tree type = builtin_types[group.ops_infos.types[vec_type_idx].index].vector; + if (strstr (group.base_name, "l") + && strstr (group.base_name, "u") + && !TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + if (strstr (group.base_name, "l") + && !strstr (group.base_name, "u") + && TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + machine_mode mode = TYPE_MODE (type); + int sew = GET_MODE_BITSIZE (GET_MODE_INNER (mode)); + if (strstr (group.base_name, "h") && sew == 8) + return; + + if (strstr (group.base_name, "w") && (sew == 8 || sew ==16)) + return; + + b.add_overloaded_function (function_instance, *group.shape); + b.add_unique_function (function_instance, (*group.shape), return_type, + argument_types); +} + +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +772,23 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1251,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1265,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..44b9fec1898 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -934,6 +934,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1481,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2672,38 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * index_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +2882,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..22fed60b4c3 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -123,6 +123,7 @@ enum required_ext ZVKNHB_EXT, /* Crypto vector Zvknhb sub-ext */ ZVKSED_EXT, /* Crypto vector Zvksed sub-ext */ ZVKSH_EXT, /* Crypto vector Zvksh sub-ext */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ }; /* Enumerates the RVV operand types. */ @@ -252,6 +253,8 @@ struct function_group_info return TARGET_ZVKSED; case ZVKSH_EXT: return TARGET_ZVKSH; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; default: gcc_unreachable (); } diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..fd3ba29bae9 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlhu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlwu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vsb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vlsb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlshu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlswu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vssb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vlxb, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxh, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxw, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxbu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxhu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxwu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vsxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..5fe9ba08c4e 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,165 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& riscv_vector::whole_reg_to_reg_move_p (operands, <MODE>mode, 7)" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..3c12c124597 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..30bef369375 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..3c8b5ccc16b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..b7c00404f18 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..17a53012acf --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..b187cfc852b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: [PATCH v6] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-12 3:22 ` [PATCH v6] RISC-V: Add support for xtheadvector-specific intrinsics Jun Sha (Joshua) @ 2024-01-12 7:32 ` juzhe.zhong 0 siblings, 0 replies; 133+ messages in thread From: juzhe.zhong @ 2024-01-12 7:32 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, cooper.joshua, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 54650 bytes --] ok juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-12 11:22 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v6] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (class th_extract): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (struct th_extract_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_FUNCTION): Redefine DEF_RVV_FUNCTION for XTheadVector special intrinsics. * config/riscv/riscv-vector-builtins.h (enum required_ext): Add new XTheadVector member. (struct function_group_info): Likewise. * config/riscv/t-riscv: Add thead-vector-builtins-functions.def * config/riscv/thead-vector.md (@pred_mov_width<vlmem_op_attr><mode>): Add new patterns. (*pred_mov_width<vlmem_op_attr><mode>): Likewise. (@pred_store_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_load_width<vlmem_op_attr><mode>): Likewise. (@pred_strided_store_width<vlmem_op_attr><mode>): Likewise. (@pred_indexed_load_width<vlmem_op_attr><mode>): Likewise. (@pred_th_extract<mode>): Likewise. (*pred_th_extract<mode>): Likewise. * config/riscv/thead-vector-builtins-functions.def: New file. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 +++ .../riscv/riscv-vector-builtins-shapes.cc | 160 +++++++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + gcc/config/riscv/riscv-vector-builtins.cc | 70 +++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 250 ++++++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 +++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 +++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 +++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 +++++ 15 files changed, 1104 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 1aa6e3c6665..b6f6e4ff37e 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2141,6 +2141,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2603,6 +2680,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> vsuxw_obj; +static CONSTEXPR const th_extract vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2894,6 +3002,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (vlb) +BASE (vlh) +BASE (vlw) +BASE (vlbu) +BASE (vlhu) +BASE (vlwu) +BASE (vsb) +BASE (vsh) +BASE (vsw) +BASE (vlsb) +BASE (vlsh) +BASE (vlsw) +BASE (vlsbu) +BASE (vlshu) +BASE (vlswu) +BASE (vssb) +BASE (vssh) +BASE (vssw) +BASE (vlxb) +BASE (vlxh) +BASE (vlxw) +BASE (vlxbu) +BASE (vlxhu) +BASE (vlxwu) +BASE (vsxb) +BASE (vsxh) +BASE (vsxw) +BASE (vsuxb) +BASE (vsuxh) +BASE (vsuxw) +BASE (vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 87c7f43a423..1f2c94d3541 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -280,6 +280,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const vlb; +extern const function_base *const vlh; +extern const function_base *const vlw; +extern const function_base *const vlbu; +extern const function_base *const vlhu; +extern const function_base *const vlwu; +extern const function_base *const vsb; +extern const function_base *const vsh; +extern const function_base *const vsw; +extern const function_base *const vlsb; +extern const function_base *const vlsh; +extern const function_base *const vlsw; +extern const function_base *const vlsbu; +extern const function_base *const vlshu; +extern const function_base *const vlswu; +extern const function_base *const vssb; +extern const function_base *const vssh; +extern const function_base *const vssw; +extern const function_base *const vlxb; +extern const function_base *const vlxh; +extern const function_base *const vlxw; +extern const function_base *const vlxbu; +extern const function_base *const vlxhu; +extern const function_base *const vlxwu; +extern const function_base *const vsxb; +extern const function_base *const vsxh; +extern const function_base *const vsxw; +extern const function_base *const vsuxb; +extern const function_base *const vsuxh; +extern const function_base *const vsuxw; +extern const function_base *const vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..8e90b17a94b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,146 @@ struct indexed_loadstore_def : public function_shape } }; +/* Add one function instance for GROUP, using operand suffix at index OI, + mode suffix at index PAIR && bi and predication suffix at index pred_idx. */ +static void +build_th_loadstore (function_builder &b, const function_group_info &group, + unsigned int pred_idx, unsigned int vec_type_idx) +{ + auto_vec<tree, 5> argument_types; + function_instance function_instance (group.base_name, *group.base, + *group.shape, + group.ops_infos.types[vec_type_idx], + group.preds[pred_idx], &group.ops_infos); + tree return_type = group.ops_infos.ret.get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + b.allocate_argument_types (function_instance, argument_types); + b.apply_predication (function_instance, return_type, argument_types); + + if (TARGET_XTHEADVECTOR && !check_type (return_type, argument_types)) + return; + + tree type = builtin_types[group.ops_infos.types[vec_type_idx].index].vector; + if (strstr (group.base_name, "l") + && strstr (group.base_name, "u") + && !TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + if (strstr (group.base_name, "l") + && !strstr (group.base_name, "u") + && TYPE_UNSIGNED (TREE_TYPE (type))) + return; + + machine_mode mode = TYPE_MODE (type); + int sew = GET_MODE_BITSIZE (GET_MODE_INNER (mode)); + if (strstr (group.base_name, "h") && sew == 8) + return; + + if (strstr (group.base_name, "w") && (sew == 8 || sew ==16)) + return; + + b.add_overloaded_function (function_instance, *group.shape); + b.add_unique_function (function_instance, (*group.shape), return_type, + argument_types); +} + +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_th_loadstore (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +772,23 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_name ("__riscv_th_"); + b.append_name (instance.base_name); + + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1251,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1265,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..44b9fec1898 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -934,6 +934,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1481,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2672,38 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_size_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * index_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_const_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info all_v_scalar_ptr_index_ops + = {all_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +2882,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..22fed60b4c3 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -123,6 +123,7 @@ enum required_ext ZVKNHB_EXT, /* Crypto vector Zvknhb sub-ext */ ZVKSED_EXT, /* Crypto vector Zvksed sub-ext */ ZVKSH_EXT, /* Crypto vector Zvksh sub-ext */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ }; /* Enumerates the RVV operand types. */ @@ -252,6 +253,8 @@ struct function_group_info return TARGET_ZVKSED; case ZVKSH_EXT: return TARGET_ZVKSH; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; default: gcc_unreachable (); } diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..fd3ba29bae9 --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlhu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vlwu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (vsb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vsw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (vlsb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsh, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsw, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlsbu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlshu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vlswu, th_loadstore_width, full_preds, all_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (vssb, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssh, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vssw, th_loadstore_width, none_m_preds, all_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (vlxb, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxh, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxw, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxbu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxhu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vlxwu, th_indexed_loadstore_width, full_preds, all_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (vsxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxb, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxh, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vsuxw, th_indexed_loadstore_width, none_m_preds, all_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..5fe9ba08c4e 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,165 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& riscv_vector::whole_reg_to_reg_move_p (operands, <MODE>mode, 7)" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..3c12c124597 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..30bef369375 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..3c8b5ccc16b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..b7c00404f18 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..17a53012acf --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..b187cfc852b --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_th_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* [PATCH v6] RISC-V: Fix register overlap issue for some xtheadvector instructions 2024-01-12 3:18 ` [PATCH v5] RISC-V: Support XTheadVector extension Jun Sha (Joshua) ` (3 preceding siblings ...) 2024-01-12 3:22 ` [PATCH v6] RISC-V: Add support for xtheadvector-specific intrinsics Jun Sha (Joshua) @ 2024-01-12 3:23 ` Jun Sha (Joshua) 2024-01-12 7:32 ` juzhe.zhong 2024-01-12 3:24 ` [PATCH v5] RISC-V: Rewrite some instructions using ASM targethook Jun Sha (Joshua) 2024-01-18 14:43 ` [PATCH v5] RISC-V: Support XTheadVector extension Christoph Müllner 6 siblings, 1 reply; 133+ messages in thread From: Jun Sha (Joshua) @ 2024-01-12 3:23 UTC (permalink / raw) To: gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, Jun Sha (Joshua), Jin Ma, Xianmiao Qu For th.vmadc/th.vmsbc as well as narrowing arithmetic instructions and floating-point compare instructions, an illegal instruction exception will be raised if the destination vector register overlaps a source vector register group. To handle this issue, we add an attribute "spec_restriction" to disable some alternatives for xtheadvector. gcc/ChangeLog: * config/riscv/riscv.md (none,thv,rvv): (no,yes): Add an attribute to disable alternative for xtheadvector or RVV1.0. * config/riscv/vector.md: Disable alternatives that destination register overlaps source register group for xtheadvector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config/riscv/riscv.md | 22 +++ gcc/config/riscv/vector.md | 314 +++++++++++++++++++++---------------- 2 files changed, 202 insertions(+), 134 deletions(-) diff --git a/gcc/config/riscv/riscv.md b/gcc/config/riscv/riscv.md index 84212430dc0..23fc32d5cb2 100644 --- a/gcc/config/riscv/riscv.md +++ b/gcc/config/riscv/riscv.md @@ -579,6 +579,25 @@ ] (const_string "yes"))) +;; This attribute marks the alternatives not matching the constraints +;; described in spec as disabled. +(define_attr "spec_restriction" "none,thv,rvv" + (const_string "none")) + +(define_attr "spec_restriction_disabled" "no,yes" + (cond [(eq_attr "spec_restriction" "none") + (const_string "no") + + (and (eq_attr "spec_restriction" "thv") + (match_test "TARGET_XTHEADVECTOR")) + (const_string "yes") + + (and (eq_attr "spec_restriction" "rvv") + (match_test "TARGET_VECTOR && !TARGET_XTHEADVECTOR")) + (const_string "yes") + ] + (const_string "no"))) + ;; Attribute to control enable or disable instructions. (define_attr "enabled" "no,yes" (cond [ @@ -590,6 +609,9 @@ (eq_attr "group_overlap_valid" "no") (const_string "no") + + (eq_attr "spec_restriction_disabled" "yes") + (const_string "no") ] (const_string "yes"))) diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index 3eb6daafbc2..c79416cf0d3 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -3260,7 +3260,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "spec_restriction" "thv,none,none")]) (define_insn "@pred_msbc<mode>" [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr") @@ -3279,7 +3280,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "spec_restriction" "thv,thv,none")]) (define_insn "@pred_madc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3299,7 +3301,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "spec_restriction" "thv,none")]) (define_insn "@pred_msbc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3319,7 +3322,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "spec_restriction" "thv,none")]) (define_expand "@pred_madc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3368,7 +3372,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "spec_restriction" "thv,none")]) (define_insn "*pred_madc<mode>_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3389,7 +3394,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "spec_restriction" "thv,none")]) (define_expand "@pred_msbc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3438,7 +3444,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "spec_restriction" "thv,none")]) (define_insn "*pred_msbc<mode>_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3459,7 +3466,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "spec_restriction" "thv,none")]) (define_insn "@pred_madc<mode>_overflow" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr, &vr") @@ -3477,7 +3485,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "spec_restriction" "thv,none,none")]) (define_insn "@pred_msbc<mode>_overflow" [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") @@ -3495,7 +3504,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "spec_restriction" "thv,thv,none,none")]) (define_insn "@pred_madc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3514,7 +3524,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "spec_restriction" "thv,none")]) (define_insn "@pred_msbc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3533,7 +3544,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "spec_restriction" "thv,none")]) (define_expand "@pred_madc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3580,7 +3592,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "spec_restriction" "thv,none")]) (define_insn "*pred_madc<mode>_overflow_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3600,7 +3613,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "spec_restriction" "thv,none")]) (define_expand "@pred_msbc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3647,7 +3661,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "spec_restriction" "thv,none")]) (define_insn "*pred_msbc<mode>_overflow_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3667,7 +3682,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "spec_restriction" "thv,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated integer unary operations @@ -3987,7 +4003,8 @@ "TARGET_VECTOR" "vn<insn>.w%o4\t%0,%3,%v4%p1" [(set_attr "type" "vnshift") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "spec_restriction" "none,none,thv,thv,none,thv,none,none,none,thv,none,none")]) (define_insn "@pred_narrow_<optab><mode>_scalar" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -4008,7 +4025,8 @@ "TARGET_VECTOR" "vn<insn>.w%o4\t%0,%3,%4%p1" [(set_attr "type" "vnshift") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "spec_restriction" "none,none,thv,thv,none,none")]) ;; vncvt.x.x.w (define_insn "@pred_trunc<mode>" @@ -4032,7 +4050,8 @@ (set_attr "vl_op_idx" "4") (set (attr "ta") (symbol_ref "riscv_vector::get_ta(operands[5])")) (set (attr "ma") (symbol_ref "riscv_vector::get_ma(operands[6])")) - (set (attr "avl_type_idx") (const_int 7))]) + (set (attr "avl_type_idx") (const_int 7)) + (set_attr "spec_restriction" "none,none,thv,thv,none,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated fixed-point operations @@ -4438,7 +4457,8 @@ "TARGET_VECTOR" "vnclip<v_su>.w%o4\t%0,%3,%v4%p1" [(set_attr "type" "vnclip") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "spec_restriction" "thv,thv,thv,thv,thv,thv,none,none,thv,thv,none,none")]) (define_insn "@pred_narrow_clip<v_su><mode>_scalar" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -4460,7 +4480,8 @@ "TARGET_VECTOR" "vnclip<v_su>.w%o4\t%0,%3,%4%p1" [(set_attr "type" "vnclip") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "spec_restriction" "thv,thv,thv,thv,none,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated integer comparison operations @@ -4511,23 +4532,24 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr, &vr, &vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") - (match_operand 7 "const_int_operand" " i, i, i, i") - (match_operand 8 "const_int_operand" " i, i, i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK, rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i, i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_ltge_operator" - [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr") - (match_operand:V_VLSI 5 "vector_arith_operand" " vr, vr, vi, vi")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] + [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr, vr, vr, vr, vr") + (match_operand:V_VLSI 5 "vector_arith_operand" " vr, vr, vi, vi, vr, vr, vi, vi")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0, vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,thv,thv,rvv,rvv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_narrow" @@ -4547,7 +4569,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,thv,thv,thv,thv,none,none")]) (define_expand "@pred_ltge<mode>" [(set (match_operand:<VM> 0 "register_operand") @@ -4591,23 +4614,24 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_ltge<mode>" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr, &vr, &vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") - (match_operand 7 "const_int_operand" " i, i, i, i") - (match_operand 8 "const_int_operand" " i, i, i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK, rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i, i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "ltge_operator" - [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr") - (match_operand:V_VLSI 5 "vector_neg_arith_operand" " vr, vr, vj, vj")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] + [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr, vr, vr, vr, vr") + (match_operand:V_VLSI 5 "vector_neg_arith_operand" " vr, vr, vj, vj, vr, vr, vj, vj")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0, vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,thv,thv,rvv,rvv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_ltge<mode>_narrow" @@ -4627,7 +4651,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,thv,thv,thv,thv,none,none")]) (define_expand "@pred_cmp<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -4673,24 +4698,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_eqge_operator" - [(match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr") + [(match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSI_QHS - (match_operand:<VEL> 5 "register_operand" " r, r"))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r"))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_scalar_narrow" @@ -4711,7 +4737,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,none,none")]) (define_expand "@pred_eqne<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -4757,24 +4784,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSI_QHS - (match_operand:<VEL> 5 "register_operand" " r, r")) - (match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r")) + (match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_eqne<mode>_scalar_narrow" @@ -4795,7 +4823,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,none,none")]) ;; Handle GET_MODE_INNER (mode) = DImode. We need to split them since ;; we need to deal with SEW = 64 in RV32 system. @@ -4922,24 +4951,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_eqge_operator" - [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr") + [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSI_D - (match_operand:<VEL> 5 "register_operand" " r, r"))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r"))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_scalar_narrow" @@ -4960,28 +4990,30 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,none,none")]) ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSI_D - (match_operand:<VEL> 5 "register_operand" " r, r")) - (match_operand:V_VLSI_D 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r")) + (match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_eqne<mode>_scalar_narrow" @@ -5002,7 +5034,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,none,none")]) (define_insn "*pred_cmp<mode>_extended_scalar_merge_tie_mask" [(set (match_operand:<VM> 0 "register_operand" "=vm") @@ -5031,25 +5064,26 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_extended_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_eqge_operator" - [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr") + [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSI_D (sign_extend:<VEL> - (match_operand:<VSUBEL> 5 "register_operand" " r, r")))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VSUBEL> 5 "register_operand" " r, r, r, r")))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) (define_insn "*pred_cmp<mode>_extended_scalar_narrow" [(set (match_operand:<VM> 0 "register_operand" "=vm, vr, vr, &vr, &vr") @@ -5070,7 +5104,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,none,none")]) (define_insn "*pred_eqne<mode>_extended_scalar_merge_tie_mask" [(set (match_operand:<VM> 0 "register_operand" "=vm") @@ -5099,25 +5134,26 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_extended_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSI_D (sign_extend:<VEL> - (match_operand:<VSUBEL> 5 "register_operand" " r, r"))) - (match_operand:V_VLSI_D 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VSUBEL> 5 "register_operand" " r, r, r, r"))) + (match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) (define_insn "*pred_eqne<mode>_extended_scalar_narrow" [(set (match_operand:<VM> 0 "register_operand" "=vm, vr, vr, &vr, &vr") @@ -5138,7 +5174,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,none,none")]) ;; GE, vmsge.vx/vmsgeu.vx ;; @@ -7327,23 +7364,24 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "signed_order_operator" - [(match_operand:V_VLSF 4 "register_operand" " vr, vr") - (match_operand:V_VLSF 5 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + [(match_operand:V_VLSF 4 "register_operand" " vr, vr, vr, vr") + (match_operand:V_VLSF 5 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vmf%B3.vv\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) (define_insn "*pred_cmp<mode>_narrow_merge_tie_mask" [(set (match_operand:<VM> 0 "register_operand" "=vm") @@ -7386,7 +7424,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vmf%B3.vv\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,thv,thv,thv,thv,none,none")]) (define_expand "@pred_cmp<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -7432,24 +7471,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "signed_order_operator" - [(match_operand:V_VLSF 4 "register_operand" " vr, vr") + [(match_operand:V_VLSF 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSF - (match_operand:<VEL> 5 "register_operand" " f, f"))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " f, f, f, f"))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_scalar_narrow" @@ -7470,7 +7510,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,none,none")]) (define_expand "@pred_eqne<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -7516,24 +7557,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSF - (match_operand:<VEL> 5 "register_operand" " f, f")) - (match_operand:V_VLSF 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " f, f, f, f")) + (match_operand:V_VLSF 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_eqne<mode>_scalar_narrow" @@ -7554,7 +7596,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,none,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated floating-point merge @@ -7774,7 +7817,8 @@ [(set_attr "type" "vfncvtftoi") (set_attr "mode" "<VNCONVERT>") (set (attr "frm_mode") - (symbol_ref "riscv_vector::get_frm_mode (operands[8])"))]) + (symbol_ref "riscv_vector::get_frm_mode (operands[8])")) + (set_attr "spec_restriction" "none,none,thv,thv,none,none")]) (define_insn "@pred_narrow_<fix_cvt><mode>" [(set (match_operand:<VNCONVERT> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -7816,7 +7860,8 @@ [(set_attr "type" "vfncvtitof") (set_attr "mode" "<VNCONVERT>") (set (attr "frm_mode") - (symbol_ref "riscv_vector::get_frm_mode (operands[8])"))]) + (symbol_ref "riscv_vector::get_frm_mode (operands[8])")) + (set_attr "spec_restriction" "none,none,thv,thv,none,none")]) (define_insn "@pred_trunc<mode>" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -7839,7 +7884,8 @@ [(set_attr "type" "vfncvtftof") (set_attr "mode" "<V_DOUBLE_TRUNC>") (set (attr "frm_mode") - (symbol_ref "riscv_vector::get_frm_mode (operands[8])"))]) + (symbol_ref "riscv_vector::get_frm_mode (operands[8])")) + (set_attr "spec_restriction" "none,none,thv,thv,none,none")]) (define_insn "@pred_rod_trunc<mode>" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: [PATCH v6] RISC-V: Fix register overlap issue for some xtheadvector instructions 2024-01-12 3:23 ` [PATCH v6] RISC-V: Fix register overlap issue for some xtheadvector instructions Jun Sha (Joshua) @ 2024-01-12 7:32 ` juzhe.zhong 0 siblings, 0 replies; 133+ messages in thread From: juzhe.zhong @ 2024-01-12 7:32 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, cooper.joshua, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 38551 bytes --] ok. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-12 11:23 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v6] RISC-V: Fix register overlap issue for some xtheadvector instructions For th.vmadc/th.vmsbc as well as narrowing arithmetic instructions and floating-point compare instructions, an illegal instruction exception will be raised if the destination vector register overlaps a source vector register group. To handle this issue, we add an attribute "spec_restriction" to disable some alternatives for xtheadvector. gcc/ChangeLog: * config/riscv/riscv.md (none,thv,rvv): (no,yes): Add an attribute to disable alternative for xtheadvector or RVV1.0. * config/riscv/vector.md: Disable alternatives that destination register overlaps source register group for xtheadvector. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config/riscv/riscv.md | 22 +++ gcc/config/riscv/vector.md | 314 +++++++++++++++++++++---------------- 2 files changed, 202 insertions(+), 134 deletions(-) diff --git a/gcc/config/riscv/riscv.md b/gcc/config/riscv/riscv.md index 84212430dc0..23fc32d5cb2 100644 --- a/gcc/config/riscv/riscv.md +++ b/gcc/config/riscv/riscv.md @@ -579,6 +579,25 @@ ] (const_string "yes"))) +;; This attribute marks the alternatives not matching the constraints +;; described in spec as disabled. +(define_attr "spec_restriction" "none,thv,rvv" + (const_string "none")) + +(define_attr "spec_restriction_disabled" "no,yes" + (cond [(eq_attr "spec_restriction" "none") + (const_string "no") + + (and (eq_attr "spec_restriction" "thv") + (match_test "TARGET_XTHEADVECTOR")) + (const_string "yes") + + (and (eq_attr "spec_restriction" "rvv") + (match_test "TARGET_VECTOR && !TARGET_XTHEADVECTOR")) + (const_string "yes") + ] + (const_string "no"))) + ;; Attribute to control enable or disable instructions. (define_attr "enabled" "no,yes" (cond [ @@ -590,6 +609,9 @@ (eq_attr "group_overlap_valid" "no") (const_string "no") + + (eq_attr "spec_restriction_disabled" "yes") + (const_string "no") ] (const_string "yes"))) diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md index 3eb6daafbc2..c79416cf0d3 100644 --- a/gcc/config/riscv/vector.md +++ b/gcc/config/riscv/vector.md @@ -3260,7 +3260,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "spec_restriction" "thv,none,none")]) (define_insn "@pred_msbc<mode>" [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr") @@ -3279,7 +3280,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "spec_restriction" "thv,thv,none")]) (define_insn "@pred_madc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3299,7 +3301,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "spec_restriction" "thv,none")]) (define_insn "@pred_msbc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3319,7 +3322,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "spec_restriction" "thv,none")]) (define_expand "@pred_madc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3368,7 +3372,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "spec_restriction" "thv,none")]) (define_insn "*pred_madc<mode>_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3389,7 +3394,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "spec_restriction" "thv,none")]) (define_expand "@pred_msbc<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3438,7 +3444,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "spec_restriction" "thv,none")]) (define_insn "*pred_msbc<mode>_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3459,7 +3466,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "4") - (set (attr "avl_type_idx") (const_int 5))]) + (set (attr "avl_type_idx") (const_int 5)) + (set_attr "spec_restriction" "thv,none")]) (define_insn "@pred_madc<mode>_overflow" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr, &vr") @@ -3477,7 +3485,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "spec_restriction" "thv,none,none")]) (define_insn "@pred_msbc<mode>_overflow" [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") @@ -3495,7 +3504,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "spec_restriction" "thv,thv,none,none")]) (define_insn "@pred_madc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3514,7 +3524,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "spec_restriction" "thv,none")]) (define_insn "@pred_msbc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3533,7 +3544,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "spec_restriction" "thv,none")]) (define_expand "@pred_madc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3580,7 +3592,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "spec_restriction" "thv,none")]) (define_insn "*pred_madc<mode>_overflow_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3600,7 +3613,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "spec_restriction" "thv,none")]) (define_expand "@pred_msbc<mode>_overflow_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -3647,7 +3661,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "spec_restriction" "thv,none")]) (define_insn "*pred_msbc<mode>_overflow_extended_scalar" [(set (match_operand:<VM> 0 "register_operand" "=vr, &vr") @@ -3667,7 +3682,8 @@ [(set_attr "type" "vicalu") (set_attr "mode" "<MODE>") (set_attr "vl_op_idx" "3") - (set (attr "avl_type_idx") (const_int 4))]) + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "spec_restriction" "thv,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated integer unary operations @@ -3987,7 +4003,8 @@ "TARGET_VECTOR" "vn<insn>.w%o4\t%0,%3,%v4%p1" [(set_attr "type" "vnshift") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "spec_restriction" "none,none,thv,thv,none,thv,none,none,none,thv,none,none")]) (define_insn "@pred_narrow_<optab><mode>_scalar" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -4008,7 +4025,8 @@ "TARGET_VECTOR" "vn<insn>.w%o4\t%0,%3,%4%p1" [(set_attr "type" "vnshift") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "spec_restriction" "none,none,thv,thv,none,none")]) ;; vncvt.x.x.w (define_insn "@pred_trunc<mode>" @@ -4032,7 +4050,8 @@ (set_attr "vl_op_idx" "4") (set (attr "ta") (symbol_ref "riscv_vector::get_ta(operands[5])")) (set (attr "ma") (symbol_ref "riscv_vector::get_ma(operands[6])")) - (set (attr "avl_type_idx") (const_int 7))]) + (set (attr "avl_type_idx") (const_int 7)) + (set_attr "spec_restriction" "none,none,thv,thv,none,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated fixed-point operations @@ -4438,7 +4457,8 @@ "TARGET_VECTOR" "vnclip<v_su>.w%o4\t%0,%3,%v4%p1" [(set_attr "type" "vnclip") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "spec_restriction" "thv,thv,thv,thv,thv,thv,none,none,thv,thv,none,none")]) (define_insn "@pred_narrow_clip<v_su><mode>_scalar" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -4460,7 +4480,8 @@ "TARGET_VECTOR" "vnclip<v_su>.w%o4\t%0,%3,%4%p1" [(set_attr "type" "vnclip") - (set_attr "mode" "<V_DOUBLE_TRUNC>")]) + (set_attr "mode" "<V_DOUBLE_TRUNC>") + (set_attr "spec_restriction" "thv,thv,thv,thv,none,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated integer comparison operations @@ -4511,23 +4532,24 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr, &vr, &vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") - (match_operand 7 "const_int_operand" " i, i, i, i") - (match_operand 8 "const_int_operand" " i, i, i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK, rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i, i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_ltge_operator" - [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr") - (match_operand:V_VLSI 5 "vector_arith_operand" " vr, vr, vi, vi")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] + [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr, vr, vr, vr, vr") + (match_operand:V_VLSI 5 "vector_arith_operand" " vr, vr, vi, vi, vr, vr, vi, vi")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0, vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,thv,thv,rvv,rvv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_narrow" @@ -4547,7 +4569,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,thv,thv,thv,thv,none,none")]) (define_expand "@pred_ltge<mode>" [(set (match_operand:<VM> 0 "register_operand") @@ -4591,23 +4614,24 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_ltge<mode>" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, vr, vr, &vr, &vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") - (match_operand 7 "const_int_operand" " i, i, i, i") - (match_operand 8 "const_int_operand" " i, i, i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK, rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i, i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "ltge_operator" - [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr") - (match_operand:V_VLSI 5 "vector_neg_arith_operand" " vr, vr, vj, vj")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] + [(match_operand:V_VLSI 4 "register_operand" " vr, vr, vr, vr, vr, vr, vr, vr") + (match_operand:V_VLSI 5 "vector_neg_arith_operand" " vr, vr, vj, vj, vr, vr, vj, vj")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0, vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,thv,thv,rvv,rvv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_ltge<mode>_narrow" @@ -4627,7 +4651,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.v%o5\t%0,%4,%v5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,thv,thv,thv,thv,none,none")]) (define_expand "@pred_cmp<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -4673,24 +4698,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_eqge_operator" - [(match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr") + [(match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSI_QHS - (match_operand:<VEL> 5 "register_operand" " r, r"))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r"))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_scalar_narrow" @@ -4711,7 +4737,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,none,none")]) (define_expand "@pred_eqne<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -4757,24 +4784,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSI_QHS - (match_operand:<VEL> 5 "register_operand" " r, r")) - (match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r")) + (match_operand:V_VLSI_QHS 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_eqne<mode>_scalar_narrow" @@ -4795,7 +4823,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,none,none")]) ;; Handle GET_MODE_INNER (mode) = DImode. We need to split them since ;; we need to deal with SEW = 64 in RV32 system. @@ -4922,24 +4951,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_eqge_operator" - [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr") + [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSI_D - (match_operand:<VEL> 5 "register_operand" " r, r"))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r"))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_scalar_narrow" @@ -4960,28 +4990,30 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,none,none")]) ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSI_D - (match_operand:<VEL> 5 "register_operand" " r, r")) - (match_operand:V_VLSI_D 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " r, r, r, r")) + (match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_eqne<mode>_scalar_narrow" @@ -5002,7 +5034,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,none,none")]) (define_insn "*pred_cmp<mode>_extended_scalar_merge_tie_mask" [(set (match_operand:<VM> 0 "register_operand" "=vm") @@ -5031,25 +5064,26 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_extended_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "comparison_except_eqge_operator" - [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr") + [(match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSI_D (sign_extend:<VEL> - (match_operand:<VSUBEL> 5 "register_operand" " r, r")))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VSUBEL> 5 "register_operand" " r, r, r, r")))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) (define_insn "*pred_cmp<mode>_extended_scalar_narrow" [(set (match_operand:<VM> 0 "register_operand" "=vm, vr, vr, &vr, &vr") @@ -5070,7 +5104,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,none,none")]) (define_insn "*pred_eqne<mode>_extended_scalar_merge_tie_mask" [(set (match_operand:<VM> 0 "register_operand" "=vm") @@ -5099,25 +5134,26 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_extended_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSI_D (sign_extend:<VEL> - (match_operand:<VSUBEL> 5 "register_operand" " r, r"))) - (match_operand:V_VLSI_D 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VSUBEL> 5 "register_operand" " r, r, r, r"))) + (match_operand:V_VLSI_D 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) (define_insn "*pred_eqne<mode>_extended_scalar_narrow" [(set (match_operand:<VM> 0 "register_operand" "=vm, vr, vr, &vr, &vr") @@ -5138,7 +5174,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode) && !TARGET_64BIT" "vms%B3.vx\t%0,%4,%5%p1" [(set_attr "type" "vicmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,none,none")]) ;; GE, vmsge.vx/vmsgeu.vx ;; @@ -7327,23 +7364,24 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "signed_order_operator" - [(match_operand:V_VLSF 4 "register_operand" " vr, vr") - (match_operand:V_VLSF 5 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + [(match_operand:V_VLSF 4 "register_operand" " vr, vr, vr, vr") + (match_operand:V_VLSF 5 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vmf%B3.vv\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) (define_insn "*pred_cmp<mode>_narrow_merge_tie_mask" [(set (match_operand:<VM> 0 "register_operand" "=vm") @@ -7386,7 +7424,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vmf%B3.vv\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,thv,thv,thv,thv,none,none")]) (define_expand "@pred_cmp<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -7432,24 +7471,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_cmp<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "signed_order_operator" - [(match_operand:V_VLSF 4 "register_operand" " vr, vr") + [(match_operand:V_VLSF 4 "register_operand" " vr, vr, vr, vr") (vec_duplicate:V_VLSF - (match_operand:<VEL> 5 "register_operand" " f, f"))]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " f, f, f, f"))]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_cmp<mode>_scalar_narrow" @@ -7470,7 +7510,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,none,none")]) (define_expand "@pred_eqne<mode>_scalar" [(set (match_operand:<VM> 0 "register_operand") @@ -7516,24 +7557,25 @@ ;; We don't use early-clobber for LMUL <= 1 to get better codegen. (define_insn "*pred_eqne<mode>_scalar" - [(set (match_operand:<VM> 0 "register_operand" "=vr, vr") + [(set (match_operand:<VM> 0 "register_operand" "=vr, vr, &vr, &vr") (if_then_else:<VM> (unspec:<VM> - [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1") - (match_operand 6 "vector_length_operand" " rK, rK") - (match_operand 7 "const_int_operand" " i, i") - (match_operand 8 "const_int_operand" " i, i") + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1,vmWc1,vmWc1,vmWc1") + (match_operand 6 "vector_length_operand" " rK, rK, rK, rK") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") (reg:SI VL_REGNUM) (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE) (match_operator:<VM> 3 "equality_operator" [(vec_duplicate:V_VLSF - (match_operand:<VEL> 5 "register_operand" " f, f")) - (match_operand:V_VLSF 4 "register_operand" " vr, vr")]) - (match_operand:<VM> 2 "vector_merge_operand" " vu, 0")))] + (match_operand:<VEL> 5 "register_operand" " f, f, f, f")) + (match_operand:V_VLSF 4 "register_operand" " vr, vr, vr, vr")]) + (match_operand:<VM> 2 "vector_merge_operand" " vu, 0, vu, 0")))] "TARGET_VECTOR && riscv_vector::cmp_lmul_le_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "thv,thv,rvv,rvv")]) ;; We use early-clobber for source LMUL > dest LMUL. (define_insn "*pred_eqne<mode>_scalar_narrow" @@ -7554,7 +7596,8 @@ "TARGET_VECTOR && riscv_vector::cmp_lmul_gt_one (<MODE>mode)" "vmf%B3.vf\t%0,%4,%5%p1" [(set_attr "type" "vfcmp") - (set_attr "mode" "<MODE>")]) + (set_attr "mode" "<MODE>") + (set_attr "spec_restriction" "none,thv,thv,none,none")]) ;; ------------------------------------------------------------------------------- ;; ---- Predicated floating-point merge @@ -7774,7 +7817,8 @@ [(set_attr "type" "vfncvtftoi") (set_attr "mode" "<VNCONVERT>") (set (attr "frm_mode") - (symbol_ref "riscv_vector::get_frm_mode (operands[8])"))]) + (symbol_ref "riscv_vector::get_frm_mode (operands[8])")) + (set_attr "spec_restriction" "none,none,thv,thv,none,none")]) (define_insn "@pred_narrow_<fix_cvt><mode>" [(set (match_operand:<VNCONVERT> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -7816,7 +7860,8 @@ [(set_attr "type" "vfncvtitof") (set_attr "mode" "<VNCONVERT>") (set (attr "frm_mode") - (symbol_ref "riscv_vector::get_frm_mode (operands[8])"))]) + (symbol_ref "riscv_vector::get_frm_mode (operands[8])")) + (set_attr "spec_restriction" "none,none,thv,thv,none,none")]) (define_insn "@pred_trunc<mode>" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") @@ -7839,7 +7884,8 @@ [(set_attr "type" "vfncvtftof") (set_attr "mode" "<V_DOUBLE_TRUNC>") (set (attr "frm_mode") - (symbol_ref "riscv_vector::get_frm_mode (operands[8])"))]) + (symbol_ref "riscv_vector::get_frm_mode (operands[8])")) + (set_attr "spec_restriction" "none,none,thv,thv,none,none")]) (define_insn "@pred_rod_trunc<mode>" [(set (match_operand:<V_DOUBLE_TRUNC> 0 "register_operand" "=vd, vd, vr, vr, &vr, &vr") -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* [PATCH v5] RISC-V: Rewrite some instructions using ASM targethook 2024-01-12 3:18 ` [PATCH v5] RISC-V: Support XTheadVector extension Jun Sha (Joshua) ` (4 preceding siblings ...) 2024-01-12 3:23 ` [PATCH v6] RISC-V: Fix register overlap issue for some xtheadvector instructions Jun Sha (Joshua) @ 2024-01-12 3:24 ` Jun Sha (Joshua) 2024-01-12 7:32 ` juzhe.zhong 2024-01-18 14:43 ` [PATCH v5] RISC-V: Support XTheadVector extension Christoph Müllner 6 siblings, 1 reply; 133+ messages in thread From: Jun Sha (Joshua) @ 2024-01-12 3:24 UTC (permalink / raw) To: gcc-patches Cc: jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, juzhe.zhong, Jun Sha (Joshua), Jin Ma, Xianmiao Qu There are some xtheadvector instructions that differ from RVV1.0 apart from simply adding "th." prefix. For example, RVV1.0 load/store instructions will have SEW while xtheadvector not; RVV1.0 will have "o" for indexed-ordered store instructions while xtheadvecotr not; xtheadvector and RVV1.0 have different vnsrl/vnsra/vfncvt suffix (vv/vx/vi vs wv/wx/wi). To address this issue without duplicating patterns, we use ASM targethook to rewrite the whole string of the instructions. We identify different instructions from the corresponding attribute. gcc/ChangeLog: * config/riscv/thead.cc (th_asm_output_opcode): Rewrite some instructions. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config/riscv/thead.cc | 215 +++++++++++++++++++++++++++++++++++++- 1 file changed, 213 insertions(+), 2 deletions(-) diff --git a/gcc/config/riscv/thead.cc b/gcc/config/riscv/thead.cc index dc3aed3904d..fb088ebff02 100644 --- a/gcc/config/riscv/thead.cc +++ b/gcc/config/riscv/thead.cc @@ -27,6 +27,7 @@ #include "backend.h" #include "tree.h" #include "rtl.h" +#include "insn-attr.h" #include "explow.h" #include "memmodel.h" #include "emit-rtl.h" @@ -890,8 +891,218 @@ th_asm_output_opcode (FILE *asm_out_file, const char *p) { /* We need to add th. prefix to all the xtheadvector instructions here.*/ - if (current_output_insn != NULL && p[0] == 'v') - fputs ("th.", asm_out_file); + if (current_output_insn != NULL) + { + if (get_attr_type (current_output_insn) == TYPE_VLDE || + get_attr_type (current_output_insn) == TYPE_VSTE || + get_attr_type (current_output_insn) == TYPE_VLDFF) + { + if (strstr (p, "e8") || strstr (p, "e16") || + strstr (p, "e32") || strstr (p, "e64")) + { + get_attr_type (current_output_insn) == TYPE_VSTE + ? fputs ("th.vse", asm_out_file) + : fputs ("th.vle", asm_out_file); + if (strstr (p, "e8")) + return p+4; + else + return p+5; + } + } + + if (get_attr_type (current_output_insn) == TYPE_VLDS || + get_attr_type (current_output_insn) == TYPE_VSTS) + { + if (strstr (p, "vle8") || strstr (p, "vse8") || + strstr (p, "vle16") || strstr (p, "vse16") || + strstr (p, "vle32") || strstr (p, "vse32") || + strstr (p, "vle64") || strstr (p, "vse64")) + { + get_attr_type (current_output_insn) == TYPE_VSTS + ? fputs ("th.vse", asm_out_file) + : fputs ("th.vle", asm_out_file); + if (strstr (p, "e8")) + return p+4; + else + return p+5; + } + else if (strstr (p, "vlse8") || strstr (p, "vsse8") || + strstr (p, "vlse16") || strstr (p, "vsse16") || + strstr (p, "vlse32") || strstr (p, "vsse32") || + strstr (p, "vlse64") || strstr (p, "vsse64")) + { + get_attr_type (current_output_insn) == TYPE_VSTS + ? fputs ("th.vsse", asm_out_file) + : fputs ("th.vlse", asm_out_file); + if (strstr (p, "e8")) + return p+5; + else + return p+6; + } + } + + if (get_attr_type (current_output_insn) == TYPE_VLDUX || + get_attr_type (current_output_insn) == TYPE_VLDOX) + { + if (strstr (p, "ei")) + { + fputs ("th.vlxe", asm_out_file); + if (strstr (p, "ei8")) + return p+7; + else + return p+8; + } + } + + if (get_attr_type (current_output_insn) == TYPE_VSTUX || + get_attr_type (current_output_insn) == TYPE_VSTOX) + { + if (strstr (p, "ei")) + { + get_attr_type (current_output_insn) == TYPE_VSTUX + ? fputs ("th.vsuxe", asm_out_file) + : fputs ("th.vsxe", asm_out_file); + if (strstr (p, "ei8")) + return p+7; + else + return p+8; + } + } + + if (get_attr_type (current_output_insn) == TYPE_VLSEGDE || + get_attr_type (current_output_insn) == TYPE_VSSEGTE || + get_attr_type (current_output_insn) == TYPE_VLSEGDFF) + { + get_attr_type (current_output_insn) == TYPE_VSSEGTE + ? fputs ("th.vsseg", asm_out_file) + : fputs ("th.vlseg", asm_out_file); + asm_fprintf (asm_out_file, "%c", p[5]); + fputs ("e", asm_out_file); + if (strstr (p, "e8")) + return p+8; + else + return p+9; + } + + if (get_attr_type (current_output_insn) == TYPE_VLSEGDS || + get_attr_type (current_output_insn) == TYPE_VSSEGTS) + { + get_attr_type (current_output_insn) == TYPE_VSSEGTS + ? fputs ("th.vssseg", asm_out_file) + : fputs ("th.vlsseg", asm_out_file); + asm_fprintf (asm_out_file, "%c", p[6]); + fputs ("e", asm_out_file); + if (strstr (p, "e8")) + return p+9; + else + return p+10; + } + + if (get_attr_type (current_output_insn) == TYPE_VLSEGDUX || + get_attr_type (current_output_insn) == TYPE_VLSEGDOX) + { + fputs ("th.vlxseg", asm_out_file); + asm_fprintf (asm_out_file, "%c", p[7]); + fputs ("e", asm_out_file); + if (strstr (p, "ei8")) + return p+11; + else + return p+12; + } + + if (get_attr_type (current_output_insn) == TYPE_VSSEGTUX || + get_attr_type (current_output_insn) == TYPE_VSSEGTOX) + { + fputs ("th.vsxseg", asm_out_file); + asm_fprintf (asm_out_file, "%c", p[7]); + fputs ("e", asm_out_file); + if (strstr (p, "ei8")) + return p+11; + else + return p+12; + } + + if (get_attr_type (current_output_insn) == TYPE_VNSHIFT) + { + if (strstr (p, "vncvt")) + { + fputs ("th.vncvt.x.x.v", asm_out_file); + return p+11; + } + + strstr (p, "vnsrl") ? fputs ("th.vnsrl.v", asm_out_file) + : fputs ("th.vnsra.v", asm_out_file); + return p+7; + } + + if (get_attr_type (current_output_insn) == TYPE_VNCLIP) + { + if (strstr (p, "vnclipu")) + { + fputs ("th.vnclipu.v", asm_out_file); + return p+9; + } + else + { + fputs ("th.vnclip.v", asm_out_file); + return p+8; + } + } + + if (get_attr_type (current_output_insn) == TYPE_VMPOP) + { + fputs ("th.vmpopc", asm_out_file); + return p+5; + } + + if (get_attr_type (current_output_insn) == TYPE_VMFFS) + { + fputs ("th.vmfirst", asm_out_file); + return p+6; + } + + if (get_attr_type (current_output_insn) == TYPE_VFNCVTFTOI || + get_attr_type (current_output_insn) == TYPE_VFNCVTITOF) + { + if (strstr (p, "xu")) + { + get_attr_type (current_output_insn) == TYPE_VFNCVTFTOI + ? fputs ("th.vfncvt.xu.f.v", asm_out_file) + : fputs ("th.vfncvt.f.xu.v", asm_out_file); + return p+13; + } + else + { + get_attr_type (current_output_insn) == TYPE_VFNCVTFTOI + ? fputs ("th.vfncvt.x.f.v", asm_out_file) + : fputs ("th.vfncvt.f.x.v", asm_out_file); + return p+12; + } + } + + if (get_attr_type (current_output_insn) == TYPE_VFNCVTFTOF) + { + fputs ("th.vfncvt.f.f.v", asm_out_file); + return p+12; + } + + if (get_attr_type (current_output_insn) == TYPE_VFREDU + && strstr (p, "sum")) + { + fputs ("th.vfredsum", asm_out_file); + return p+9; + } + + if (get_attr_type (current_output_insn) == TYPE_VFWREDU + && strstr (p, "sum")) + { + fputs ("th.vfwredsum", asm_out_file); + return p+10; + } + + if (p[0] == 'v') + fputs ("th.", asm_out_file); + } return p; } -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: [PATCH v5] RISC-V: Rewrite some instructions using ASM targethook 2024-01-12 3:24 ` [PATCH v5] RISC-V: Rewrite some instructions using ASM targethook Jun Sha (Joshua) @ 2024-01-12 7:32 ` juzhe.zhong 0 siblings, 0 replies; 133+ messages in thread From: juzhe.zhong @ 2024-01-12 7:32 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, cooper.joshua, jinma, cooper.qu [-- Attachment #1: Type: text/plain, Size: 7778 bytes --] ok. juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-12 11:24 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Rewrite some instructions using ASM targethook There are some xtheadvector instructions that differ from RVV1.0 apart from simply adding "th." prefix. For example, RVV1.0 load/store instructions will have SEW while xtheadvector not; RVV1.0 will have "o" for indexed-ordered store instructions while xtheadvecotr not; xtheadvector and RVV1.0 have different vnsrl/vnsra/vfncvt suffix (vv/vx/vi vs wv/wx/wi). To address this issue without duplicating patterns, we use ASM targethook to rewrite the whole string of the instructions. We identify different instructions from the corresponding attribute. gcc/ChangeLog: * config/riscv/thead.cc (th_asm_output_opcode): Rewrite some instructions. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- gcc/config/riscv/thead.cc | 215 +++++++++++++++++++++++++++++++++++++- 1 file changed, 213 insertions(+), 2 deletions(-) diff --git a/gcc/config/riscv/thead.cc b/gcc/config/riscv/thead.cc index dc3aed3904d..fb088ebff02 100644 --- a/gcc/config/riscv/thead.cc +++ b/gcc/config/riscv/thead.cc @@ -27,6 +27,7 @@ #include "backend.h" #include "tree.h" #include "rtl.h" +#include "insn-attr.h" #include "explow.h" #include "memmodel.h" #include "emit-rtl.h" @@ -890,8 +891,218 @@ th_asm_output_opcode (FILE *asm_out_file, const char *p) { /* We need to add th. prefix to all the xtheadvector instructions here.*/ - if (current_output_insn != NULL && p[0] == 'v') - fputs ("th.", asm_out_file); + if (current_output_insn != NULL) + { + if (get_attr_type (current_output_insn) == TYPE_VLDE || + get_attr_type (current_output_insn) == TYPE_VSTE || + get_attr_type (current_output_insn) == TYPE_VLDFF) + { + if (strstr (p, "e8") || strstr (p, "e16") || + strstr (p, "e32") || strstr (p, "e64")) + { + get_attr_type (current_output_insn) == TYPE_VSTE + ? fputs ("th.vse", asm_out_file) + : fputs ("th.vle", asm_out_file); + if (strstr (p, "e8")) + return p+4; + else + return p+5; + } + } + + if (get_attr_type (current_output_insn) == TYPE_VLDS || + get_attr_type (current_output_insn) == TYPE_VSTS) + { + if (strstr (p, "vle8") || strstr (p, "vse8") || + strstr (p, "vle16") || strstr (p, "vse16") || + strstr (p, "vle32") || strstr (p, "vse32") || + strstr (p, "vle64") || strstr (p, "vse64")) + { + get_attr_type (current_output_insn) == TYPE_VSTS + ? fputs ("th.vse", asm_out_file) + : fputs ("th.vle", asm_out_file); + if (strstr (p, "e8")) + return p+4; + else + return p+5; + } + else if (strstr (p, "vlse8") || strstr (p, "vsse8") || + strstr (p, "vlse16") || strstr (p, "vsse16") || + strstr (p, "vlse32") || strstr (p, "vsse32") || + strstr (p, "vlse64") || strstr (p, "vsse64")) + { + get_attr_type (current_output_insn) == TYPE_VSTS + ? fputs ("th.vsse", asm_out_file) + : fputs ("th.vlse", asm_out_file); + if (strstr (p, "e8")) + return p+5; + else + return p+6; + } + } + + if (get_attr_type (current_output_insn) == TYPE_VLDUX || + get_attr_type (current_output_insn) == TYPE_VLDOX) + { + if (strstr (p, "ei")) + { + fputs ("th.vlxe", asm_out_file); + if (strstr (p, "ei8")) + return p+7; + else + return p+8; + } + } + + if (get_attr_type (current_output_insn) == TYPE_VSTUX || + get_attr_type (current_output_insn) == TYPE_VSTOX) + { + if (strstr (p, "ei")) + { + get_attr_type (current_output_insn) == TYPE_VSTUX + ? fputs ("th.vsuxe", asm_out_file) + : fputs ("th.vsxe", asm_out_file); + if (strstr (p, "ei8")) + return p+7; + else + return p+8; + } + } + + if (get_attr_type (current_output_insn) == TYPE_VLSEGDE || + get_attr_type (current_output_insn) == TYPE_VSSEGTE || + get_attr_type (current_output_insn) == TYPE_VLSEGDFF) + { + get_attr_type (current_output_insn) == TYPE_VSSEGTE + ? fputs ("th.vsseg", asm_out_file) + : fputs ("th.vlseg", asm_out_file); + asm_fprintf (asm_out_file, "%c", p[5]); + fputs ("e", asm_out_file); + if (strstr (p, "e8")) + return p+8; + else + return p+9; + } + + if (get_attr_type (current_output_insn) == TYPE_VLSEGDS || + get_attr_type (current_output_insn) == TYPE_VSSEGTS) + { + get_attr_type (current_output_insn) == TYPE_VSSEGTS + ? fputs ("th.vssseg", asm_out_file) + : fputs ("th.vlsseg", asm_out_file); + asm_fprintf (asm_out_file, "%c", p[6]); + fputs ("e", asm_out_file); + if (strstr (p, "e8")) + return p+9; + else + return p+10; + } + + if (get_attr_type (current_output_insn) == TYPE_VLSEGDUX || + get_attr_type (current_output_insn) == TYPE_VLSEGDOX) + { + fputs ("th.vlxseg", asm_out_file); + asm_fprintf (asm_out_file, "%c", p[7]); + fputs ("e", asm_out_file); + if (strstr (p, "ei8")) + return p+11; + else + return p+12; + } + + if (get_attr_type (current_output_insn) == TYPE_VSSEGTUX || + get_attr_type (current_output_insn) == TYPE_VSSEGTOX) + { + fputs ("th.vsxseg", asm_out_file); + asm_fprintf (asm_out_file, "%c", p[7]); + fputs ("e", asm_out_file); + if (strstr (p, "ei8")) + return p+11; + else + return p+12; + } + + if (get_attr_type (current_output_insn) == TYPE_VNSHIFT) + { + if (strstr (p, "vncvt")) + { + fputs ("th.vncvt.x.x.v", asm_out_file); + return p+11; + } + + strstr (p, "vnsrl") ? fputs ("th.vnsrl.v", asm_out_file) + : fputs ("th.vnsra.v", asm_out_file); + return p+7; + } + + if (get_attr_type (current_output_insn) == TYPE_VNCLIP) + { + if (strstr (p, "vnclipu")) + { + fputs ("th.vnclipu.v", asm_out_file); + return p+9; + } + else + { + fputs ("th.vnclip.v", asm_out_file); + return p+8; + } + } + + if (get_attr_type (current_output_insn) == TYPE_VMPOP) + { + fputs ("th.vmpopc", asm_out_file); + return p+5; + } + + if (get_attr_type (current_output_insn) == TYPE_VMFFS) + { + fputs ("th.vmfirst", asm_out_file); + return p+6; + } + + if (get_attr_type (current_output_insn) == TYPE_VFNCVTFTOI || + get_attr_type (current_output_insn) == TYPE_VFNCVTITOF) + { + if (strstr (p, "xu")) + { + get_attr_type (current_output_insn) == TYPE_VFNCVTFTOI + ? fputs ("th.vfncvt.xu.f.v", asm_out_file) + : fputs ("th.vfncvt.f.xu.v", asm_out_file); + return p+13; + } + else + { + get_attr_type (current_output_insn) == TYPE_VFNCVTFTOI + ? fputs ("th.vfncvt.x.f.v", asm_out_file) + : fputs ("th.vfncvt.f.x.v", asm_out_file); + return p+12; + } + } + + if (get_attr_type (current_output_insn) == TYPE_VFNCVTFTOF) + { + fputs ("th.vfncvt.f.f.v", asm_out_file); + return p+12; + } + + if (get_attr_type (current_output_insn) == TYPE_VFREDU + && strstr (p, "sum")) + { + fputs ("th.vfredsum", asm_out_file); + return p+9; + } + + if (get_attr_type (current_output_insn) == TYPE_VFWREDU + && strstr (p, "sum")) + { + fputs ("th.vfwredsum", asm_out_file); + return p+10; + } + + if (p[0] == 'v') + fputs ("th.", asm_out_file); + } return p; } -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: [PATCH v5] RISC-V: Support XTheadVector extension 2024-01-12 3:18 ` [PATCH v5] RISC-V: Support XTheadVector extension Jun Sha (Joshua) ` (5 preceding siblings ...) 2024-01-12 3:24 ` [PATCH v5] RISC-V: Rewrite some instructions using ASM targethook Jun Sha (Joshua) @ 2024-01-18 14:43 ` Christoph Müllner 2024-01-19 20:03 ` Jeff Law 6 siblings, 1 reply; 133+ messages in thread From: Christoph Müllner @ 2024-01-18 14:43 UTC (permalink / raw) To: Jun Sha (Joshua) Cc: gcc-patches, jim.wilson.gcc, palmer, andrew, philipp.tomsich, jeffreyalaw, juzhe.zhong, kito.cheng, Jin Ma, Xianmiao Qu On Fri, Jan 12, 2024 at 4:18 AM Jun Sha (Joshua) <cooper.joshua@linux.alibaba.com> wrote: > > This patch series presents gcc implementation of the XTheadVector > extension [1]. > > [1] https://github.com/T-head-Semi/thead-extension-spec/ > > For some vector patterns that cannot be avoided, we use > "!TARGET_XTHEADVECTOR" to disable them in order not to > generate instructions that xtheadvector does not support, > causing 10 changes in vector.md. > > For the th. prefix issue, we use current_output_insn and > the ASM_OUTPUT_OPCODE hook instead of directly modifying > patterns in vector.md. > > We have run the GCC test suite and can confirm that there > are no regressions. > > Furthermore, we have run the tests in > https://github.com/riscv-non-isa/rvv-intrinsic-doc/tree/main/examples, > and all the tests passed. > > Co-authored-by: Jin Ma <jinma@linux.alibaba.com> > Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> > Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> > > [PATCH v4] RISC-V: Introduce XTheadVector as a subset of V1.0.0 > [PATCH v5] RISC-V: Adds the prefix "th." for the instructions of XTheadVector > [PATCH v6] RISC-V: Handle differences between XTheadvector and Vector > [PATCH v6] RISC-V: Add support for xtheadvector-specific intrinsics > [PATCH v6] RISC-V: Fix register overlap issue for some xtheadvector instructions > [PATCH v5] RISC-V: Rewrite some instructions using ASM targethook All patches of this series got either "LGTM" or "OK": * https://gcc.gnu.org/pipermail/gcc-patches/2024-January/643339.html * https://gcc.gnu.org/pipermail/gcc-patches/2024-January/642798.html * https://gcc.gnu.org/pipermail/gcc-patches/2024-January/642799.html * https://gcc.gnu.org/pipermail/gcc-patches/2024-January/642800.html * https://gcc.gnu.org/pipermail/gcc-patches/2024-January/642801.html * https://gcc.gnu.org/pipermail/gcc-patches/2024-January/642802.html As mentioned earlier, I have rebased the patches, retested them locally and (after ensuring there are no regressions) pushed them. To all involved people: thank you very much! A special 'thank you' goes to Juzhe, who did a great job in reviewing the patches and providing suggestions to get the code into shape! ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: [PATCH v5] RISC-V: Support XTheadVector extension 2024-01-18 14:43 ` [PATCH v5] RISC-V: Support XTheadVector extension Christoph Müllner @ 2024-01-19 20:03 ` Jeff Law 0 siblings, 0 replies; 133+ messages in thread From: Jeff Law @ 2024-01-19 20:03 UTC (permalink / raw) To: Christoph Müllner, Jun Sha (Joshua) Cc: gcc-patches, jim.wilson.gcc, palmer, andrew, philipp.tomsich, juzhe.zhong, kito.cheng, Jin Ma, Xianmiao Qu On 1/18/24 07:43, Christoph Müllner wrote: > On Fri, Jan 12, 2024 at 4:18 AM Jun Sha (Joshua) > <cooper.joshua@linux.alibaba.com> wrote: >> >> This patch series presents gcc implementation of the XTheadVector >> extension [1]. >> >> [1] https://github.com/T-head-Semi/thead-extension-spec/ >> >> For some vector patterns that cannot be avoided, we use >> "!TARGET_XTHEADVECTOR" to disable them in order not to >> generate instructions that xtheadvector does not support, >> causing 10 changes in vector.md. >> >> For the th. prefix issue, we use current_output_insn and >> the ASM_OUTPUT_OPCODE hook instead of directly modifying >> patterns in vector.md. >> >> We have run the GCC test suite and can confirm that there >> are no regressions. >> >> Furthermore, we have run the tests in >> https://github.com/riscv-non-isa/rvv-intrinsic-doc/tree/main/examples, >> and all the tests passed. >> >> Co-authored-by: Jin Ma <jinma@linux.alibaba.com> >> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> >> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> >> >> [PATCH v4] RISC-V: Introduce XTheadVector as a subset of V1.0.0 >> [PATCH v5] RISC-V: Adds the prefix "th." for the instructions of XTheadVector >> [PATCH v6] RISC-V: Handle differences between XTheadvector and Vector >> [PATCH v6] RISC-V: Add support for xtheadvector-specific intrinsics >> [PATCH v6] RISC-V: Fix register overlap issue for some xtheadvector instructions >> [PATCH v5] RISC-V: Rewrite some instructions using ASM targethook > > All patches of this series got either "LGTM" or "OK": > * https://gcc.gnu.org/pipermail/gcc-patches/2024-January/643339.html > * https://gcc.gnu.org/pipermail/gcc-patches/2024-January/642798.html > * https://gcc.gnu.org/pipermail/gcc-patches/2024-January/642799.html > * https://gcc.gnu.org/pipermail/gcc-patches/2024-January/642800.html > * https://gcc.gnu.org/pipermail/gcc-patches/2024-January/642801.html > * https://gcc.gnu.org/pipermail/gcc-patches/2024-January/642802.html > > As mentioned earlier, I have rebased the patches, retested them locally and > (after ensuring there are no regressions) pushed them. > > To all involved people: thank you very much! > A special 'thank you' goes to Juzhe, who did a great job in reviewing > the patches > and providing suggestions to get the code into shape! Likewise. Glad to see we were able to push this through to a reasonable conclusion and a huge thanks to Juzhe for all his work on the review side. Jeff ^ permalink raw reply [flat|nested] 133+ messages in thread
[parent not found: <84F7577214BA46AD+202401101900269053910@rivai.ai>]
* Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. [not found] <84F7577214BA46AD+202401101900269053910@rivai.ai> @ 2024-01-10 11:06 ` joshua 2024-01-10 11:08 ` joshua 0 siblings, 1 reply; 133+ messages in thread From: joshua @ 2024-01-10 11:06 UTC (permalink / raw) To: juzhe.zhong, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu The key difference between vlb/vlh/vlw is not output type too. Their difference is the range of datatype, not one specific type. We have dived into the xtheadvector special intrinsics and are sure about that. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月10日(星期三) 19:00 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. instance.op_info->args[i].get_tree_type (instance.type.index) is output type. You can use GDB debug it . juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-10 18:57 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Hi Juzhe, Perhaps things are not as simple as imagined. The differences between vlb/vlh/vlw is not the same as vle8/vle16/vle32. "8", "16" or "32" in vle8/vle16/vle32 can be appended from "vle" according to input type. But vlb/vlh/vlw is different not in input type. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月10日(星期三) 18:03 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. I mean change these: +DEF_RVV_FUNCTION (th_vlb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_ops) into a single: +DEF_RVV_FUNCTION (th_vl, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) and append "h", "w", or"b" according to TYPE_UNSIGNED and GET_MODE_BITSIZE (GET_MODE_INNER (TYPE_MODE (instance.op_info->args[i].get_tree_type (instance.type.index)))) in th_loadstore_width. It should definitely works, I allow this flexibility in design of the framework. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-10 17:55 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. And revise th_loadstore_width, append the name according TYPE_UNSIGNED and GET_MODE_BITSIZE (GET_MODE_INNER (TYPE_MODE (instance.op_info->args[i].get_tree_type (instance.type.index)))) What do you mean by it? I'm a bit confused. Changing i8_v_scalar_const_ptr_ops into all_v_scalar_const_ptr_ops will expand the datatypes that can be used in th_vlb. Can we restrict again in th_loadstore_width? ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月10日(星期三) 17:35 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. +DEF_RVV_FUNCTION (th_vlb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_ops) I think we should remove those many data structure you added like: i8_v_scalar_const_ptr_ops Instead, you should use all_v_scalar_const_ptr_ops And revise th_loadstore_width, append the name according TYPE_UNSIGNED and GET_MODE_BITSIZE (GET_MODE_INNER (TYPE_MODE (instance.op_info->args[i].get_tree_type (instance.type.index)))) juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-10 17:27 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-functions.def (vlsegff): Include thead-vector-builtins-functions.def. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins-types.def (DEF_RVV_I8_OPS): Add datatypes for XTheadVector. (DEF_RVV_I16_OPS): Add datatypes for XTheadVector. (DEF_RVV_I32_OPS): Add datatypes for XTheadVector. (DEF_RVV_U8_OPS): Add datatypes for XTheadVector. (DEF_RVV_U16_OPS): Add datatypes for XTheadVector. (DEF_RVV_U32_OPS): Add datatypes for XTheadVector. (vint8m1_t): Add datatypes for XTheadVector. (vint8m2_t): Likewise. (vint8m4_t): Likewise. (vint8m8_t): Likewise. (vint16m1_t): Likewise. (vint16m2_t): Likewise. (vint16m4_t): Likewise. (vint16m8_t): Likewise. (vint32m1_t): Likewise. (vint32m2_t): Likewise. (vint32m4_t): Likewise. (vint32m8_t): Likewise. (vint64m1_t): Likewise. (vint64m2_t): Likewise. (vint64m4_t): Likewise. (vint64m8_t): Likewise. (vuint8m1_t): Likewise. (vuint8m2_t): Likewise. (vuint8m4_t): Likewise. (vuint8m8_t): Likewise. (vuint16m1_t): Likewise. (vuint16m2_t): Likewise. (vuint16m4_t): Likewise. (vuint16m8_t): Likewise. (vuint32m1_t): Likewise. (vuint32m2_t): Likewise. (vuint32m4_t): Likewise. (vuint32m8_t): Likewise. (vuint64m1_t): Likewise. (vuint64m2_t): Likewise. (vuint64m4_t): Likewise. (vuint64m8_t): Likewise. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_I8_OPS): Add datatypes for XTheadVector. (DEF_RVV_I16_OPS): Add datatypes for XTheadVector. (DEF_RVV_I32_OPS): Add datatypes for XTheadVector. (DEF_RVV_U8_OPS): Add datatypes for XTheadVector. (DEF_RVV_U16_OPS): Add datatypes for XTheadVector. (DEF_RVV_U32_OPS): Add datatypes for XTheadVector. * config/riscv/thead-vector-builtins-functions.def: New file. * config/riscv/thead-vector.md: Add new patterns. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 ++ .../riscv/riscv-vector-builtins-shapes.cc | 98 ++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + .../riscv/riscv-vector-builtins-types.def | 120 +++++++ gcc/config/riscv/riscv-vector-builtins.cc | 311 ++++++++++++++++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 253 ++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 ++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 ++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 ++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 ++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 ++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 ++++ 16 files changed, 1406 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 46f1a1da33e..5f44f31a12b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2125,6 +2125,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2587,6 +2664,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> th_vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> th_vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> th_vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> th_vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> th_vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> th_vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> th_vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> th_vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> th_vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> th_vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> th_vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> th_vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> th_vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> th_vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> th_vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> th_vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> th_vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> th_vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> th_vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> th_vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> th_vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> th_vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> th_vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> th_vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> th_vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> th_vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> th_vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> th_vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> th_vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> th_vsuxw_obj; +static CONSTEXPR const th_extract th_vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2878,6 +2986,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (th_vlb) +BASE (th_vlh) +BASE (th_vlw) +BASE (th_vlbu) +BASE (th_vlhu) +BASE (th_vlwu) +BASE (th_vsb) +BASE (th_vsh) +BASE (th_vsw) +BASE (th_vlsb) +BASE (th_vlsh) +BASE (th_vlsw) +BASE (th_vlsbu) +BASE (th_vlshu) +BASE (th_vlswu) +BASE (th_vssb) +BASE (th_vssh) +BASE (th_vssw) +BASE (th_vlxb) +BASE (th_vlxh) +BASE (th_vlxw) +BASE (th_vlxbu) +BASE (th_vlxhu) +BASE (th_vlxwu) +BASE (th_vsxb) +BASE (th_vsxh) +BASE (th_vsxw) +BASE (th_vsuxb) +BASE (th_vsuxh) +BASE (th_vsuxw) +BASE (th_vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 1122e3801a7..df43adf9a17 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -299,6 +299,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const th_vlb; +extern const function_base *const th_vlh; +extern const function_base *const th_vlw; +extern const function_base *const th_vlbu; +extern const function_base *const th_vlhu; +extern const function_base *const th_vlwu; +extern const function_base *const th_vsb; +extern const function_base *const th_vsh; +extern const function_base *const th_vsw; +extern const function_base *const th_vlsb; +extern const function_base *const th_vlsh; +extern const function_base *const th_vlsw; +extern const function_base *const th_vlsbu; +extern const function_base *const th_vlshu; +extern const function_base *const th_vlswu; +extern const function_base *const th_vssb; +extern const function_base *const th_vssh; +extern const function_base *const th_vssw; +extern const function_base *const th_vlxb; +extern const function_base *const th_vlxh; +extern const function_base *const th_vlxw; +extern const function_base *const th_vlxbu; +extern const function_base *const th_vlxhu; +extern const function_base *const th_vlxwu; +extern const function_base *const th_vsxb; +extern const function_base *const th_vsxh; +extern const function_base *const th_vsxw; +extern const function_base *const th_vsuxb; +extern const function_base *const th_vsuxh; +extern const function_base *const th_vsuxw; +extern const function_base *const th_vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..489a95cf684 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,86 @@ struct indexed_loadstore_def : public function_shape } }; +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_base_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_one (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_base_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +712,21 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_base_name (instance.base_name); + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1189,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1203,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins-types.def b/gcc/config/riscv/riscv-vector-builtins-types.def index 61019a56844..abfeb4fcd9b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-types.def +++ b/gcc/config/riscv/riscv-vector-builtins-types.def @@ -24,12 +24,48 @@ along with GCC; see the file COPYING3. If not see #define DEF_RVV_I_OPS(TYPE, REQUIRE) #endif +/* Use "DEF_RVV_I8_OPS" macro include some signed integer (i8/i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I8_OPS +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_I16_OPS" macro include some signed integer (i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I16_OPS +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_I32_OPS" macro include some signed integer (i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I32_OPS +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) +#endif + /* Use "DEF_RVV_U_OPS" macro include all unsigned integer which will be iterated and registered as intrinsic functions. */ #ifndef DEF_RVV_U_OPS #define DEF_RVV_U_OPS(TYPE, REQUIRE) #endif +/* Use "DEF_RVV_U8_OPS" macro include some unsigned integer (i8/i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U8_OPS +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_U16_OPS" macro include some unsigned integer (i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U16_OPS +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_U32_OPS" macro include some unsigned integer (i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U32_OPS +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) +#endif + /* Use "DEF_RVV_F_OPS" macro include all floating-point which will be iterated and registered as intrinsic functions. */ #ifndef DEF_RVV_F_OPS @@ -374,6 +410,45 @@ DEF_RVV_I_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) DEF_RVV_I_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) DEF_RVV_I_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint8m1_t, 0) +DEF_RVV_I8_OPS (vint8m2_t, 0) +DEF_RVV_I8_OPS (vint8m4_t, 0) +DEF_RVV_I8_OPS (vint8m8_t, 0) +DEF_RVV_I8_OPS (vint16m1_t, 0) +DEF_RVV_I8_OPS (vint16m2_t, 0) +DEF_RVV_I8_OPS (vint16m4_t, 0) +DEF_RVV_I8_OPS (vint16m8_t, 0) +DEF_RVV_I8_OPS (vint32m1_t, 0) +DEF_RVV_I8_OPS (vint32m2_t, 0) +DEF_RVV_I8_OPS (vint32m4_t, 0) +DEF_RVV_I8_OPS (vint32m8_t, 0) +DEF_RVV_I8_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_I16_OPS (vint16m1_t, 0) +DEF_RVV_I16_OPS (vint16m2_t, 0) +DEF_RVV_I16_OPS (vint16m4_t, 0) +DEF_RVV_I16_OPS (vint16m8_t, 0) +DEF_RVV_I16_OPS (vint32m1_t, 0) +DEF_RVV_I16_OPS (vint32m2_t, 0) +DEF_RVV_I16_OPS (vint32m4_t, 0) +DEF_RVV_I16_OPS (vint32m8_t, 0) +DEF_RVV_I16_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_I32_OPS (vint32m1_t, 0) +DEF_RVV_I32_OPS (vint32m2_t, 0) +DEF_RVV_I32_OPS (vint32m4_t, 0) +DEF_RVV_I32_OPS (vint32m8_t, 0) +DEF_RVV_I32_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + DEF_RVV_U_OPS (vuint8mf8_t, RVV_REQUIRE_MIN_VLEN_64) DEF_RVV_U_OPS (vuint8mf4_t, 0) DEF_RVV_U_OPS (vuint8mf2_t, 0) @@ -397,6 +472,45 @@ DEF_RVV_U_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) DEF_RVV_U_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) DEF_RVV_U_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint8m1_t, 0) +DEF_RVV_U8_OPS (vuint8m2_t, 0) +DEF_RVV_U8_OPS (vuint8m4_t, 0) +DEF_RVV_U8_OPS (vuint8m8_t, 0) +DEF_RVV_U8_OPS (vuint16m1_t, 0) +DEF_RVV_U8_OPS (vuint16m2_t, 0) +DEF_RVV_U8_OPS (vuint16m4_t, 0) +DEF_RVV_U8_OPS (vuint16m8_t, 0) +DEF_RVV_U8_OPS (vuint32m1_t, 0) +DEF_RVV_U8_OPS (vuint32m2_t, 0) +DEF_RVV_U8_OPS (vuint32m4_t, 0) +DEF_RVV_U8_OPS (vuint32m8_t, 0) +DEF_RVV_U8_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_U16_OPS (vuint16m1_t, 0) +DEF_RVV_U16_OPS (vuint16m2_t, 0) +DEF_RVV_U16_OPS (vuint16m4_t, 0) +DEF_RVV_U16_OPS (vuint16m8_t, 0) +DEF_RVV_U16_OPS (vuint32m1_t, 0) +DEF_RVV_U16_OPS (vuint32m2_t, 0) +DEF_RVV_U16_OPS (vuint32m4_t, 0) +DEF_RVV_U16_OPS (vuint32m8_t, 0) +DEF_RVV_U16_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_U32_OPS (vuint32m1_t, 0) +DEF_RVV_U32_OPS (vuint32m2_t, 0) +DEF_RVV_U32_OPS (vuint32m4_t, 0) +DEF_RVV_U32_OPS (vuint32m8_t, 0) +DEF_RVV_U32_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + DEF_RVV_F_OPS (vfloat16mf4_t, RVV_REQUIRE_ELEN_FP_16 | RVV_REQUIRE_MIN_VLEN_64) DEF_RVV_F_OPS (vfloat16mf2_t, RVV_REQUIRE_ELEN_FP_16) DEF_RVV_F_OPS (vfloat16m1_t, RVV_REQUIRE_ELEN_FP_16) @@ -1379,7 +1493,13 @@ DEF_RVV_CRYPTO_SEW64_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) DEF_RVV_CRYPTO_SEW64_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) #undef DEF_RVV_I_OPS +#undef DEF_RVV_I8_OPS +#undef DEF_RVV_I16_OPS +#undef DEF_RVV_I32_OPS #undef DEF_RVV_U_OPS +#undef DEF_RVV_U8_OPS +#undef DEF_RVV_U16_OPS +#undef DEF_RVV_U32_OPS #undef DEF_RVV_F_OPS #undef DEF_RVV_B_OPS #undef DEF_RVV_WEXTI_OPS diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..f429f12dc18 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -246,6 +246,63 @@ static const rvv_type_info iu_ops[] = { #include "riscv-vector-builtins-types.def" {NUM_VECTOR_TYPES, 0}}; +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i8_ops[] = { +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i16_ops[] = { +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i32_ops[] = { +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u8_ops[] = { +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u16_ops[] = { +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u32_ops[] = { +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu8_ops[] = { +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu16_ops[] = { +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu32_ops[] = { +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + /* A list of all types will be registered for intrinsic functions. */ static const rvv_type_info all_ops[] = { #define DEF_RVV_I_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, @@ -934,6 +991,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1538,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2729,222 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_size_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_size_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_size_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_size_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_size_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_size_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_index_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_index_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_index_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_index_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_index_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_index_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew8_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_index_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew16_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_index_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew32_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_index_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type, + * function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type) + * function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type) + * function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_size_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_size_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_size_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +3123,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..a8ee39a3cb2 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -114,6 +114,7 @@ static const unsigned int CP_WRITE_CSR = 1U << 5; enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ @@ -234,6 +235,8 @@ struct function_group_info { case VECTOR_EXT: return TARGET_VECTOR; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; case ZVBB_EXT: return TARGET_ZVBB; case ZVBB_OR_ZVKB_EXT: diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..667820d4c3e --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (th_vlb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlbu, th_loadstore_width, full_preds, u8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlhu, th_loadstore_width, full_preds, u16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlwu, th_loadstore_width, full_preds, u32_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vsb, th_loadstore_width, none_m_preds, iu8_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vsh, th_loadstore_width, none_m_preds, iu16_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vsw, th_loadstore_width, none_m_preds, iu32_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vlsb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsbu, th_loadstore_width, full_preds, u8_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlshu, th_loadstore_width, full_preds, u16_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlswu, th_loadstore_width, full_preds, u32_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssb, th_loadstore_width, none_m_preds, iu8_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssh, th_loadstore_width, none_m_preds, iu16_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssw, th_loadstore_width, none_m_preds, iu32_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlxb, th_indexed_loadstore_width, full_preds, i8_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxh, th_indexed_loadstore_width, full_preds, i16_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxw, th_indexed_loadstore_width, full_preds, i32_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxbu, th_indexed_loadstore_width, full_preds, u8_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxhu, th_indexed_loadstore_width, full_preds, u16_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxwu, th_indexed_loadstore_width, full_preds, u32_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxb, th_indexed_loadstore_width, none_m_preds, iu8_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxh, th_indexed_loadstore_width, none_m_preds, iu16_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxw, th_indexed_loadstore_width, none_m_preds, iu32_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxb, th_indexed_loadstore_width, none_m_preds, iu8_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxh, th_indexed_loadstore_width, none_m_preds, iu16_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxw, th_indexed_loadstore_width, none_m_preds, iu32_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..0f3700d9269 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,168 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..4e192bbf025 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..1538afec68e --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..bf4924a1d76 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..8c451845175 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..0f5b09684a5 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..aaa75be023d --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} \ No newline at end of file -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-10 11:06 ` Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics joshua @ 2024-01-10 11:08 ` joshua 0 siblings, 0 replies; 133+ messages in thread From: joshua @ 2024-01-10 11:08 UTC (permalink / raw) To: joshua, juzhe.zhong, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu Can you see the images that I sent to you in the last email？ If not, maybe you can refer to the last chapter in the thead spec. ------------------------------------------------------------------ 发件人：joshua <cooper.joshua@linux.alibaba.com> 发送时间：2024年1月10日(星期三) 19:06 收件人："juzhe.zhong@rivai.ai"<juzhe.zhong@rivai.ai>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. The key difference between vlb/vlh/vlw is not output type too. Their difference is the range of datatype, not one specific type. We have dived into the xtheadvector special intrinsics and are sure about that. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月10日(星期三) 19:00 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. instance.op_info->args[i].get_tree_type (instance.type.index) is output type. You can use GDB debug it . juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-10 18:57 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Hi Juzhe, Perhaps things are not as simple as imagined. The differences between vlb/vlh/vlw is not the same as vle8/vle16/vle32. "8", "16" or "32" in vle8/vle16/vle32 can be appended from "vle" according to input type. But vlb/vlh/vlw is different not in input type. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月10日(星期三) 18:03 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. I mean change these: +DEF_RVV_FUNCTION (th_vlb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_ops) into a single: +DEF_RVV_FUNCTION (th_vl, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) and append "h", "w", or"b" according to TYPE_UNSIGNED and GET_MODE_BITSIZE (GET_MODE_INNER (TYPE_MODE (instance.op_info->args[i].get_tree_type (instance.type.index)))) in th_loadstore_width. It should definitely works, I allow this flexibility in design of the framework. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-10 17:55 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. And revise th_loadstore_width, append the name according TYPE_UNSIGNED and GET_MODE_BITSIZE (GET_MODE_INNER (TYPE_MODE (instance.op_info->args[i].get_tree_type (instance.type.index)))) What do you mean by it? I'm a bit confused. Changing i8_v_scalar_const_ptr_ops into all_v_scalar_const_ptr_ops will expand the datatypes that can be used in th_vlb. Can we restrict again in th_loadstore_width? ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月10日(星期三) 17:35 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. +DEF_RVV_FUNCTION (th_vlb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_ops) I think we should remove those many data structure you added like: i8_v_scalar_const_ptr_ops Instead, you should use all_v_scalar_const_ptr_ops And revise th_loadstore_width, append the name according TYPE_UNSIGNED and GET_MODE_BITSIZE (GET_MODE_INNER (TYPE_MODE (instance.op_info->args[i].get_tree_type (instance.type.index)))) juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-10 17:27 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-functions.def (vlsegff): Include thead-vector-builtins-functions.def. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins-types.def (DEF_RVV_I8_OPS): Add datatypes for XTheadVector. (DEF_RVV_I16_OPS): Add datatypes for XTheadVector. (DEF_RVV_I32_OPS): Add datatypes for XTheadVector. (DEF_RVV_U8_OPS): Add datatypes for XTheadVector. (DEF_RVV_U16_OPS): Add datatypes for XTheadVector. (DEF_RVV_U32_OPS): Add datatypes for XTheadVector. (vint8m1_t): Add datatypes for XTheadVector. (vint8m2_t): Likewise. (vint8m4_t): Likewise. (vint8m8_t): Likewise. (vint16m1_t): Likewise. (vint16m2_t): Likewise. (vint16m4_t): Likewise. (vint16m8_t): Likewise. (vint32m1_t): Likewise. (vint32m2_t): Likewise. (vint32m4_t): Likewise. (vint32m8_t): Likewise. (vint64m1_t): Likewise. (vint64m2_t): Likewise. (vint64m4_t): Likewise. (vint64m8_t): Likewise. (vuint8m1_t): Likewise. (vuint8m2_t): Likewise. (vuint8m4_t): Likewise. (vuint8m8_t): Likewise. (vuint16m1_t): Likewise. (vuint16m2_t): Likewise. (vuint16m4_t): Likewise. (vuint16m8_t): Likewise. (vuint32m1_t): Likewise. (vuint32m2_t): Likewise. (vuint32m4_t): Likewise. (vuint32m8_t): Likewise. (vuint64m1_t): Likewise. (vuint64m2_t): Likewise. (vuint64m4_t): Likewise. (vuint64m8_t): Likewise. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_I8_OPS): Add datatypes for XTheadVector. (DEF_RVV_I16_OPS): Add datatypes for XTheadVector. (DEF_RVV_I32_OPS): Add datatypes for XTheadVector. (DEF_RVV_U8_OPS): Add datatypes for XTheadVector. (DEF_RVV_U16_OPS): Add datatypes for XTheadVector. (DEF_RVV_U32_OPS): Add datatypes for XTheadVector. * config/riscv/thead-vector-builtins-functions.def: New file. * config/riscv/thead-vector.md: Add new patterns. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 ++ .../riscv/riscv-vector-builtins-shapes.cc | 98 ++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + .../riscv/riscv-vector-builtins-types.def | 120 +++++++ gcc/config/riscv/riscv-vector-builtins.cc | 311 ++++++++++++++++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 253 ++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 ++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 ++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 ++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 ++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 ++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 ++++ 16 files changed, 1406 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 46f1a1da33e..5f44f31a12b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2125,6 +2125,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2587,6 +2664,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> th_vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> th_vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> th_vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> th_vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> th_vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> th_vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> th_vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> th_vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> th_vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> th_vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> th_vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> th_vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> th_vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> th_vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> th_vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> th_vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> th_vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> th_vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> th_vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> th_vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> th_vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> th_vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> th_vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> th_vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> th_vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> th_vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> th_vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> th_vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> th_vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> th_vsuxw_obj; +static CONSTEXPR const th_extract th_vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2878,6 +2986,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (th_vlb) +BASE (th_vlh) +BASE (th_vlw) +BASE (th_vlbu) +BASE (th_vlhu) +BASE (th_vlwu) +BASE (th_vsb) +BASE (th_vsh) +BASE (th_vsw) +BASE (th_vlsb) +BASE (th_vlsh) +BASE (th_vlsw) +BASE (th_vlsbu) +BASE (th_vlshu) +BASE (th_vlswu) +BASE (th_vssb) +BASE (th_vssh) +BASE (th_vssw) +BASE (th_vlxb) +BASE (th_vlxh) +BASE (th_vlxw) +BASE (th_vlxbu) +BASE (th_vlxhu) +BASE (th_vlxwu) +BASE (th_vsxb) +BASE (th_vsxh) +BASE (th_vsxw) +BASE (th_vsuxb) +BASE (th_vsuxh) +BASE (th_vsuxw) +BASE (th_vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 1122e3801a7..df43adf9a17 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -299,6 +299,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const th_vlb; +extern const function_base *const th_vlh; +extern const function_base *const th_vlw; +extern const function_base *const th_vlbu; +extern const function_base *const th_vlhu; +extern const function_base *const th_vlwu; +extern const function_base *const th_vsb; +extern const function_base *const th_vsh; +extern const function_base *const th_vsw; +extern const function_base *const th_vlsb; +extern const function_base *const th_vlsh; +extern const function_base *const th_vlsw; +extern const function_base *const th_vlsbu; +extern const function_base *const th_vlshu; +extern const function_base *const th_vlswu; +extern const function_base *const th_vssb; +extern const function_base *const th_vssh; +extern const function_base *const th_vssw; +extern const function_base *const th_vlxb; +extern const function_base *const th_vlxh; +extern const function_base *const th_vlxw; +extern const function_base *const th_vlxbu; +extern const function_base *const th_vlxhu; +extern const function_base *const th_vlxwu; +extern const function_base *const th_vsxb; +extern const function_base *const th_vsxh; +extern const function_base *const th_vsxw; +extern const function_base *const th_vsuxb; +extern const function_base *const th_vsuxh; +extern const function_base *const th_vsuxw; +extern const function_base *const th_vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..489a95cf684 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,86 @@ struct indexed_loadstore_def : public function_shape } }; +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_base_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_one (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_base_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +712,21 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_base_name (instance.base_name); + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1189,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1203,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins-types.def b/gcc/config/riscv/riscv-vector-builtins-types.def index 61019a56844..abfeb4fcd9b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-types.def +++ b/gcc/config/riscv/riscv-vector-builtins-types.def @@ -24,12 +24,48 @@ along with GCC; see the file COPYING3. If not see #define DEF_RVV_I_OPS(TYPE, REQUIRE) #endif +/* Use "DEF_RVV_I8_OPS" macro include some signed integer (i8/i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I8_OPS +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_I16_OPS" macro include some signed integer (i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I16_OPS +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_I32_OPS" macro include some signed integer (i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I32_OPS +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) +#endif + /* Use "DEF_RVV_U_OPS" macro include all unsigned integer which will be iterated and registered as intrinsic functions. */ #ifndef DEF_RVV_U_OPS #define DEF_RVV_U_OPS(TYPE, REQUIRE) #endif +/* Use "DEF_RVV_U8_OPS" macro include some unsigned integer (i8/i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U8_OPS +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_U16_OPS" macro include some unsigned integer (i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U16_OPS +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_U32_OPS" macro include some unsigned integer (i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U32_OPS +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) +#endif + /* Use "DEF_RVV_F_OPS" macro include all floating-point which will be iterated and registered as intrinsic functions. */ #ifndef DEF_RVV_F_OPS @@ -374,6 +410,45 @@ DEF_RVV_I_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) DEF_RVV_I_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) DEF_RVV_I_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint8m1_t, 0) +DEF_RVV_I8_OPS (vint8m2_t, 0) +DEF_RVV_I8_OPS (vint8m4_t, 0) +DEF_RVV_I8_OPS (vint8m8_t, 0) +DEF_RVV_I8_OPS (vint16m1_t, 0) +DEF_RVV_I8_OPS (vint16m2_t, 0) +DEF_RVV_I8_OPS (vint16m4_t, 0) +DEF_RVV_I8_OPS (vint16m8_t, 0) +DEF_RVV_I8_OPS (vint32m1_t, 0) +DEF_RVV_I8_OPS (vint32m2_t, 0) +DEF_RVV_I8_OPS (vint32m4_t, 0) +DEF_RVV_I8_OPS (vint32m8_t, 0) +DEF_RVV_I8_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_I16_OPS (vint16m1_t, 0) +DEF_RVV_I16_OPS (vint16m2_t, 0) +DEF_RVV_I16_OPS (vint16m4_t, 0) +DEF_RVV_I16_OPS (vint16m8_t, 0) +DEF_RVV_I16_OPS (vint32m1_t, 0) +DEF_RVV_I16_OPS (vint32m2_t, 0) +DEF_RVV_I16_OPS (vint32m4_t, 0) +DEF_RVV_I16_OPS (vint32m8_t, 0) +DEF_RVV_I16_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_I32_OPS (vint32m1_t, 0) +DEF_RVV_I32_OPS (vint32m2_t, 0) +DEF_RVV_I32_OPS (vint32m4_t, 0) +DEF_RVV_I32_OPS (vint32m8_t, 0) +DEF_RVV_I32_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + DEF_RVV_U_OPS (vuint8mf8_t, RVV_REQUIRE_MIN_VLEN_64) DEF_RVV_U_OPS (vuint8mf4_t, 0) DEF_RVV_U_OPS (vuint8mf2_t, 0) @@ -397,6 +472,45 @@ DEF_RVV_U_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) DEF_RVV_U_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) DEF_RVV_U_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint8m1_t, 0) +DEF_RVV_U8_OPS (vuint8m2_t, 0) +DEF_RVV_U8_OPS (vuint8m4_t, 0) +DEF_RVV_U8_OPS (vuint8m8_t, 0) +DEF_RVV_U8_OPS (vuint16m1_t, 0) +DEF_RVV_U8_OPS (vuint16m2_t, 0) +DEF_RVV_U8_OPS (vuint16m4_t, 0) +DEF_RVV_U8_OPS (vuint16m8_t, 0) +DEF_RVV_U8_OPS (vuint32m1_t, 0) +DEF_RVV_U8_OPS (vuint32m2_t, 0) +DEF_RVV_U8_OPS (vuint32m4_t, 0) +DEF_RVV_U8_OPS (vuint32m8_t, 0) +DEF_RVV_U8_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_U16_OPS (vuint16m1_t, 0) +DEF_RVV_U16_OPS (vuint16m2_t, 0) +DEF_RVV_U16_OPS (vuint16m4_t, 0) +DEF_RVV_U16_OPS (vuint16m8_t, 0) +DEF_RVV_U16_OPS (vuint32m1_t, 0) +DEF_RVV_U16_OPS (vuint32m2_t, 0) +DEF_RVV_U16_OPS (vuint32m4_t, 0) +DEF_RVV_U16_OPS (vuint32m8_t, 0) +DEF_RVV_U16_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_U32_OPS (vuint32m1_t, 0) +DEF_RVV_U32_OPS (vuint32m2_t, 0) +DEF_RVV_U32_OPS (vuint32m4_t, 0) +DEF_RVV_U32_OPS (vuint32m8_t, 0) +DEF_RVV_U32_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + DEF_RVV_F_OPS (vfloat16mf4_t, RVV_REQUIRE_ELEN_FP_16 | RVV_REQUIRE_MIN_VLEN_64) DEF_RVV_F_OPS (vfloat16mf2_t, RVV_REQUIRE_ELEN_FP_16) DEF_RVV_F_OPS (vfloat16m1_t, RVV_REQUIRE_ELEN_FP_16) @@ -1379,7 +1493,13 @@ DEF_RVV_CRYPTO_SEW64_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) DEF_RVV_CRYPTO_SEW64_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) #undef DEF_RVV_I_OPS +#undef DEF_RVV_I8_OPS +#undef DEF_RVV_I16_OPS +#undef DEF_RVV_I32_OPS #undef DEF_RVV_U_OPS +#undef DEF_RVV_U8_OPS +#undef DEF_RVV_U16_OPS +#undef DEF_RVV_U32_OPS #undef DEF_RVV_F_OPS #undef DEF_RVV_B_OPS #undef DEF_RVV_WEXTI_OPS diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..f429f12dc18 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -246,6 +246,63 @@ static const rvv_type_info iu_ops[] = { #include "riscv-vector-builtins-types.def" {NUM_VECTOR_TYPES, 0}}; +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i8_ops[] = { +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i16_ops[] = { +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i32_ops[] = { +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u8_ops[] = { +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u16_ops[] = { +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u32_ops[] = { +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu8_ops[] = { +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu16_ops[] = { +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu32_ops[] = { +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + /* A list of all types will be registered for intrinsic functions. */ static const rvv_type_info all_ops[] = { #define DEF_RVV_I_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, @@ -934,6 +991,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1538,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2729,222 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_size_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_size_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_size_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_size_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_size_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_size_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_index_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_index_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_index_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_index_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_index_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_index_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew8_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_index_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew16_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_index_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew32_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_index_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type, + * function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type) + * function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type) + * function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_size_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_size_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_size_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +3123,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..a8ee39a3cb2 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -114,6 +114,7 @@ static const unsigned int CP_WRITE_CSR = 1U << 5; enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ @@ -234,6 +235,8 @@ struct function_group_info { case VECTOR_EXT: return TARGET_VECTOR; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; case ZVBB_EXT: return TARGET_ZVBB; case ZVBB_OR_ZVKB_EXT: diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..667820d4c3e --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (th_vlb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlbu, th_loadstore_width, full_preds, u8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlhu, th_loadstore_width, full_preds, u16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlwu, th_loadstore_width, full_preds, u32_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vsb, th_loadstore_width, none_m_preds, iu8_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vsh, th_loadstore_width, none_m_preds, iu16_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vsw, th_loadstore_width, none_m_preds, iu32_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vlsb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsbu, th_loadstore_width, full_preds, u8_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlshu, th_loadstore_width, full_preds, u16_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlswu, th_loadstore_width, full_preds, u32_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssb, th_loadstore_width, none_m_preds, iu8_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssh, th_loadstore_width, none_m_preds, iu16_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssw, th_loadstore_width, none_m_preds, iu32_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlxb, th_indexed_loadstore_width, full_preds, i8_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxh, th_indexed_loadstore_width, full_preds, i16_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxw, th_indexed_loadstore_width, full_preds, i32_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxbu, th_indexed_loadstore_width, full_preds, u8_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxhu, th_indexed_loadstore_width, full_preds, u16_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxwu, th_indexed_loadstore_width, full_preds, u32_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxb, th_indexed_loadstore_width, none_m_preds, iu8_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxh, th_indexed_loadstore_width, none_m_preds, iu16_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxw, th_indexed_loadstore_width, none_m_preds, iu32_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxb, th_indexed_loadstore_width, none_m_preds, iu8_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxh, th_indexed_loadstore_width, none_m_preds, iu16_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxw, th_indexed_loadstore_width, none_m_preds, iu32_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..0f3700d9269 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,168 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..4e192bbf025 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..1538afec68e --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..bf4924a1d76 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..8c451845175 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..0f5b09684a5 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..aaa75be023d --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} \ No newline at end of file -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-10 11:06 ` Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics joshua @ 2024-01-10 11:09 juzhe.zhong 2024-01-10 11:14 ` Re：Re：[PATCH " joshua 0 siblings, 1 reply; 133+ messages in thread From: juzhe.zhong @ 2024-01-10 11:09 UTC (permalink / raw) To: cooper.joshua, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu So vlb has not only sew = 8 ? But why do you add intrinsics as follows ? +DEF_RVV_FUNCTION (th_vlb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_ops) Why it is not : DEF_RVV_FUNCTION (th_vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) ? juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-10 19:06 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. The key difference between vlb/vlh/vlw is not output type too. Their difference is the range of datatype, not one specific type. We have dived into the xtheadvector special intrinsics and are sure about that. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月10日(星期三) 19:00 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. instance.op_info->args[i].get_tree_type (instance.type.index) is output type. You can use GDB debug it . juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-10 18:57 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Hi Juzhe, Perhaps things are not as simple as imagined. The differences between vlb/vlh/vlw is not the same as vle8/vle16/vle32. "8", "16" or "32" in vle8/vle16/vle32 can be appended from "vle" according to input type. But vlb/vlh/vlw is different not in input type. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月10日(星期三) 18:03 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. I mean change these: +DEF_RVV_FUNCTION (th_vlb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_ops) into a single: +DEF_RVV_FUNCTION (th_vl, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) and append "h", "w", or"b" according to TYPE_UNSIGNED and GET_MODE_BITSIZE (GET_MODE_INNER (TYPE_MODE (instance.op_info->args[i].get_tree_type (instance.type.index)))) in th_loadstore_width. It should definitely works, I allow this flexibility in design of the framework. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-10 17:55 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. And revise th_loadstore_width, append the name according TYPE_UNSIGNED and GET_MODE_BITSIZE (GET_MODE_INNER (TYPE_MODE (instance.op_info->args[i].get_tree_type (instance.type.index)))) What do you mean by it? I'm a bit confused. Changing i8_v_scalar_const_ptr_ops into all_v_scalar_const_ptr_ops will expand the datatypes that can be used in th_vlb. Can we restrict again in th_loadstore_width? ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月10日(星期三) 17:35 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. +DEF_RVV_FUNCTION (th_vlb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_ops) I think we should remove those many data structure you added like: i8_v_scalar_const_ptr_ops Instead, you should use all_v_scalar_const_ptr_ops And revise th_loadstore_width, append the name according TYPE_UNSIGNED and GET_MODE_BITSIZE (GET_MODE_INNER (TYPE_MODE (instance.op_info->args[i].get_tree_type (instance.type.index)))) juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-10 17:27 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-functions.def (vlsegff): Include thead-vector-builtins-functions.def. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins-types.def (DEF_RVV_I8_OPS): Add datatypes for XTheadVector. (DEF_RVV_I16_OPS): Add datatypes for XTheadVector. (DEF_RVV_I32_OPS): Add datatypes for XTheadVector. (DEF_RVV_U8_OPS): Add datatypes for XTheadVector. (DEF_RVV_U16_OPS): Add datatypes for XTheadVector. (DEF_RVV_U32_OPS): Add datatypes for XTheadVector. (vint8m1_t): Add datatypes for XTheadVector. (vint8m2_t): Likewise. (vint8m4_t): Likewise. (vint8m8_t): Likewise. (vint16m1_t): Likewise. (vint16m2_t): Likewise. (vint16m4_t): Likewise. (vint16m8_t): Likewise. (vint32m1_t): Likewise. (vint32m2_t): Likewise. (vint32m4_t): Likewise. (vint32m8_t): Likewise. (vint64m1_t): Likewise. (vint64m2_t): Likewise. (vint64m4_t): Likewise. (vint64m8_t): Likewise. (vuint8m1_t): Likewise. (vuint8m2_t): Likewise. (vuint8m4_t): Likewise. (vuint8m8_t): Likewise. (vuint16m1_t): Likewise. (vuint16m2_t): Likewise. (vuint16m4_t): Likewise. (vuint16m8_t): Likewise. (vuint32m1_t): Likewise. (vuint32m2_t): Likewise. (vuint32m4_t): Likewise. (vuint32m8_t): Likewise. (vuint64m1_t): Likewise. (vuint64m2_t): Likewise. (vuint64m4_t): Likewise. (vuint64m8_t): Likewise. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_I8_OPS): Add datatypes for XTheadVector. (DEF_RVV_I16_OPS): Add datatypes for XTheadVector. (DEF_RVV_I32_OPS): Add datatypes for XTheadVector. (DEF_RVV_U8_OPS): Add datatypes for XTheadVector. (DEF_RVV_U16_OPS): Add datatypes for XTheadVector. (DEF_RVV_U32_OPS): Add datatypes for XTheadVector. * config/riscv/thead-vector-builtins-functions.def: New file. * config/riscv/thead-vector.md: Add new patterns. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 ++ .../riscv/riscv-vector-builtins-shapes.cc | 98 ++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + .../riscv/riscv-vector-builtins-types.def | 120 +++++++ gcc/config/riscv/riscv-vector-builtins.cc | 311 ++++++++++++++++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 253 ++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 ++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 ++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 ++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 ++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 ++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 ++++ 16 files changed, 1406 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 46f1a1da33e..5f44f31a12b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2125,6 +2125,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2587,6 +2664,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> th_vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> th_vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> th_vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> th_vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> th_vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> th_vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> th_vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> th_vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> th_vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> th_vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> th_vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> th_vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> th_vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> th_vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> th_vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> th_vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> th_vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> th_vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> th_vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> th_vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> th_vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> th_vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> th_vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> th_vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> th_vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> th_vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> th_vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> th_vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> th_vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> th_vsuxw_obj; +static CONSTEXPR const th_extract th_vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2878,6 +2986,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (th_vlb) +BASE (th_vlh) +BASE (th_vlw) +BASE (th_vlbu) +BASE (th_vlhu) +BASE (th_vlwu) +BASE (th_vsb) +BASE (th_vsh) +BASE (th_vsw) +BASE (th_vlsb) +BASE (th_vlsh) +BASE (th_vlsw) +BASE (th_vlsbu) +BASE (th_vlshu) +BASE (th_vlswu) +BASE (th_vssb) +BASE (th_vssh) +BASE (th_vssw) +BASE (th_vlxb) +BASE (th_vlxh) +BASE (th_vlxw) +BASE (th_vlxbu) +BASE (th_vlxhu) +BASE (th_vlxwu) +BASE (th_vsxb) +BASE (th_vsxh) +BASE (th_vsxw) +BASE (th_vsuxb) +BASE (th_vsuxh) +BASE (th_vsuxw) +BASE (th_vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 1122e3801a7..df43adf9a17 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -299,6 +299,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const th_vlb; +extern const function_base *const th_vlh; +extern const function_base *const th_vlw; +extern const function_base *const th_vlbu; +extern const function_base *const th_vlhu; +extern const function_base *const th_vlwu; +extern const function_base *const th_vsb; +extern const function_base *const th_vsh; +extern const function_base *const th_vsw; +extern const function_base *const th_vlsb; +extern const function_base *const th_vlsh; +extern const function_base *const th_vlsw; +extern const function_base *const th_vlsbu; +extern const function_base *const th_vlshu; +extern const function_base *const th_vlswu; +extern const function_base *const th_vssb; +extern const function_base *const th_vssh; +extern const function_base *const th_vssw; +extern const function_base *const th_vlxb; +extern const function_base *const th_vlxh; +extern const function_base *const th_vlxw; +extern const function_base *const th_vlxbu; +extern const function_base *const th_vlxhu; +extern const function_base *const th_vlxwu; +extern const function_base *const th_vsxb; +extern const function_base *const th_vsxh; +extern const function_base *const th_vsxw; +extern const function_base *const th_vsuxb; +extern const function_base *const th_vsuxh; +extern const function_base *const th_vsuxw; +extern const function_base *const th_vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..489a95cf684 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,86 @@ struct indexed_loadstore_def : public function_shape } }; +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_base_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_one (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_base_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +712,21 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_base_name (instance.base_name); + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1189,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1203,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins-types.def b/gcc/config/riscv/riscv-vector-builtins-types.def index 61019a56844..abfeb4fcd9b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-types.def +++ b/gcc/config/riscv/riscv-vector-builtins-types.def @@ -24,12 +24,48 @@ along with GCC; see the file COPYING3. If not see #define DEF_RVV_I_OPS(TYPE, REQUIRE) #endif +/* Use "DEF_RVV_I8_OPS" macro include some signed integer (i8/i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I8_OPS +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_I16_OPS" macro include some signed integer (i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I16_OPS +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_I32_OPS" macro include some signed integer (i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I32_OPS +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) +#endif + /* Use "DEF_RVV_U_OPS" macro include all unsigned integer which will be iterated and registered as intrinsic functions. */ #ifndef DEF_RVV_U_OPS #define DEF_RVV_U_OPS(TYPE, REQUIRE) #endif +/* Use "DEF_RVV_U8_OPS" macro include some unsigned integer (i8/i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U8_OPS +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_U16_OPS" macro include some unsigned integer (i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U16_OPS +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_U32_OPS" macro include some unsigned integer (i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U32_OPS +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) +#endif + /* Use "DEF_RVV_F_OPS" macro include all floating-point which will be iterated and registered as intrinsic functions. */ #ifndef DEF_RVV_F_OPS @@ -374,6 +410,45 @@ DEF_RVV_I_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) DEF_RVV_I_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) DEF_RVV_I_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint8m1_t, 0) +DEF_RVV_I8_OPS (vint8m2_t, 0) +DEF_RVV_I8_OPS (vint8m4_t, 0) +DEF_RVV_I8_OPS (vint8m8_t, 0) +DEF_RVV_I8_OPS (vint16m1_t, 0) +DEF_RVV_I8_OPS (vint16m2_t, 0) +DEF_RVV_I8_OPS (vint16m4_t, 0) +DEF_RVV_I8_OPS (vint16m8_t, 0) +DEF_RVV_I8_OPS (vint32m1_t, 0) +DEF_RVV_I8_OPS (vint32m2_t, 0) +DEF_RVV_I8_OPS (vint32m4_t, 0) +DEF_RVV_I8_OPS (vint32m8_t, 0) +DEF_RVV_I8_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_I16_OPS (vint16m1_t, 0) +DEF_RVV_I16_OPS (vint16m2_t, 0) +DEF_RVV_I16_OPS (vint16m4_t, 0) +DEF_RVV_I16_OPS (vint16m8_t, 0) +DEF_RVV_I16_OPS (vint32m1_t, 0) +DEF_RVV_I16_OPS (vint32m2_t, 0) +DEF_RVV_I16_OPS (vint32m4_t, 0) +DEF_RVV_I16_OPS (vint32m8_t, 0) +DEF_RVV_I16_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_I32_OPS (vint32m1_t, 0) +DEF_RVV_I32_OPS (vint32m2_t, 0) +DEF_RVV_I32_OPS (vint32m4_t, 0) +DEF_RVV_I32_OPS (vint32m8_t, 0) +DEF_RVV_I32_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + DEF_RVV_U_OPS (vuint8mf8_t, RVV_REQUIRE_MIN_VLEN_64) DEF_RVV_U_OPS (vuint8mf4_t, 0) DEF_RVV_U_OPS (vuint8mf2_t, 0) @@ -397,6 +472,45 @@ DEF_RVV_U_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) DEF_RVV_U_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) DEF_RVV_U_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint8m1_t, 0) +DEF_RVV_U8_OPS (vuint8m2_t, 0) +DEF_RVV_U8_OPS (vuint8m4_t, 0) +DEF_RVV_U8_OPS (vuint8m8_t, 0) +DEF_RVV_U8_OPS (vuint16m1_t, 0) +DEF_RVV_U8_OPS (vuint16m2_t, 0) +DEF_RVV_U8_OPS (vuint16m4_t, 0) +DEF_RVV_U8_OPS (vuint16m8_t, 0) +DEF_RVV_U8_OPS (vuint32m1_t, 0) +DEF_RVV_U8_OPS (vuint32m2_t, 0) +DEF_RVV_U8_OPS (vuint32m4_t, 0) +DEF_RVV_U8_OPS (vuint32m8_t, 0) +DEF_RVV_U8_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_U16_OPS (vuint16m1_t, 0) +DEF_RVV_U16_OPS (vuint16m2_t, 0) +DEF_RVV_U16_OPS (vuint16m4_t, 0) +DEF_RVV_U16_OPS (vuint16m8_t, 0) +DEF_RVV_U16_OPS (vuint32m1_t, 0) +DEF_RVV_U16_OPS (vuint32m2_t, 0) +DEF_RVV_U16_OPS (vuint32m4_t, 0) +DEF_RVV_U16_OPS (vuint32m8_t, 0) +DEF_RVV_U16_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_U32_OPS (vuint32m1_t, 0) +DEF_RVV_U32_OPS (vuint32m2_t, 0) +DEF_RVV_U32_OPS (vuint32m4_t, 0) +DEF_RVV_U32_OPS (vuint32m8_t, 0) +DEF_RVV_U32_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + DEF_RVV_F_OPS (vfloat16mf4_t, RVV_REQUIRE_ELEN_FP_16 | RVV_REQUIRE_MIN_VLEN_64) DEF_RVV_F_OPS (vfloat16mf2_t, RVV_REQUIRE_ELEN_FP_16) DEF_RVV_F_OPS (vfloat16m1_t, RVV_REQUIRE_ELEN_FP_16) @@ -1379,7 +1493,13 @@ DEF_RVV_CRYPTO_SEW64_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) DEF_RVV_CRYPTO_SEW64_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) #undef DEF_RVV_I_OPS +#undef DEF_RVV_I8_OPS +#undef DEF_RVV_I16_OPS +#undef DEF_RVV_I32_OPS #undef DEF_RVV_U_OPS +#undef DEF_RVV_U8_OPS +#undef DEF_RVV_U16_OPS +#undef DEF_RVV_U32_OPS #undef DEF_RVV_F_OPS #undef DEF_RVV_B_OPS #undef DEF_RVV_WEXTI_OPS diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..f429f12dc18 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -246,6 +246,63 @@ static const rvv_type_info iu_ops[] = { #include "riscv-vector-builtins-types.def" {NUM_VECTOR_TYPES, 0}}; +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i8_ops[] = { +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i16_ops[] = { +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i32_ops[] = { +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u8_ops[] = { +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u16_ops[] = { +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u32_ops[] = { +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu8_ops[] = { +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu16_ops[] = { +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu32_ops[] = { +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + /* A list of all types will be registered for intrinsic functions. */ static const rvv_type_info all_ops[] = { #define DEF_RVV_I_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, @@ -934,6 +991,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1538,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2729,222 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_size_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_size_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_size_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_size_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_size_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_size_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_index_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_index_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_index_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_index_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_index_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_index_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew8_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_index_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew16_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_index_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew32_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_index_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type, + * function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type) + * function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type) + * function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_size_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_size_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_size_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +3123,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..a8ee39a3cb2 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -114,6 +114,7 @@ static const unsigned int CP_WRITE_CSR = 1U << 5; enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ @@ -234,6 +235,8 @@ struct function_group_info { case VECTOR_EXT: return TARGET_VECTOR; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; case ZVBB_EXT: return TARGET_ZVBB; case ZVBB_OR_ZVKB_EXT: diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..667820d4c3e --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (th_vlb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlbu, th_loadstore_width, full_preds, u8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlhu, th_loadstore_width, full_preds, u16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlwu, th_loadstore_width, full_preds, u32_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vsb, th_loadstore_width, none_m_preds, iu8_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vsh, th_loadstore_width, none_m_preds, iu16_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vsw, th_loadstore_width, none_m_preds, iu32_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vlsb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsbu, th_loadstore_width, full_preds, u8_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlshu, th_loadstore_width, full_preds, u16_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlswu, th_loadstore_width, full_preds, u32_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssb, th_loadstore_width, none_m_preds, iu8_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssh, th_loadstore_width, none_m_preds, iu16_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssw, th_loadstore_width, none_m_preds, iu32_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlxb, th_indexed_loadstore_width, full_preds, i8_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxh, th_indexed_loadstore_width, full_preds, i16_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxw, th_indexed_loadstore_width, full_preds, i32_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxbu, th_indexed_loadstore_width, full_preds, u8_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxhu, th_indexed_loadstore_width, full_preds, u16_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxwu, th_indexed_loadstore_width, full_preds, u32_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxb, th_indexed_loadstore_width, none_m_preds, iu8_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxh, th_indexed_loadstore_width, none_m_preds, iu16_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxw, th_indexed_loadstore_width, none_m_preds, iu32_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxb, th_indexed_loadstore_width, none_m_preds, iu8_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxh, th_indexed_loadstore_width, none_m_preds, iu16_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxw, th_indexed_loadstore_width, none_m_preds, iu32_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..0f3700d9269 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,168 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..4e192bbf025 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..1538afec68e --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..bf4924a1d76 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..8c451845175 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..0f5b09684a5 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..aaa75be023d --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} \ No newline at end of file -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
* Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. 2024-01-10 11:09 Re：[PATCH " juzhe.zhong @ 2024-01-10 11:14 ` joshua 0 siblings, 0 replies; 133+ messages in thread From: joshua @ 2024-01-10 11:14 UTC (permalink / raw) To: juzhe.zhong, gcc-patches Cc: Jim Wilson, palmer, andrew, philipp.tomsich, jeffreyalaw, christoph.muellner, jinma, cooper.qu vlb can accept sew=8/16/32/64. vlh can accept sew=16/32/64; vlw can accept sew=32/64. vint8m1_t __riscv_th_vlb_v_i8m1 (const int8_t *a, size_t vl); vint8m2_t __riscv_th_vlb_v_i8m2 (const int8_t *a, size_t vl); vint8m4_t __riscv_th_vlb_v_i8m4 (const int8_t *a, size_t vl); vint8m8_t __riscv_th_vlb_v_i8m8 (const int8_t *a, size_t vl); vint16m1_t __riscv_th_vlb_v_i16m1 (const int16_t *a, size_t vl); vint16m2_t __riscv_th_vlb_v_i16m2 (const int16_t *a, size_t vl); vint16m4_t __riscv_th_vlb_v_i16m4 (const int16_t *a, size_t vl); vint16m8_t __riscv_th_vlb_v_i16m8 (const int16_t *a, size_t vl); vint32m1_t __riscv_th_vlb_v_i32m1 (const int32_t *a, size_t vl); vint32m2_t __riscv_th_vlb_v_i32m2 (const int32_t *a, size_t vl); vint32m4_t __riscv_th_vlb_v_i32m4 (const int32_t *a, size_t vl); vint32m8_t __riscv_th_vlb_v_i32m8 (const int32_t *a, size_t vl); vint64m1_t __riscv_th_vlb_v_i64m1 (const int64_t *a, size_t vl); vint64m2_t __riscv_th_vlb_v_i64m2 (const int64_t *a, size_t vl); vint64m4_t __riscv_th_vlb_v_i64m4 (const int64_t *a, size_t vl); vint64m8_t __riscv_th_vlb_v_i64m8 (const int64_t *a, size_t vl); vint16m1_t __riscv_th_vlh_v_i16m1 (const int16_t *a, size_t vl); vint16m2_t __riscv_th_vlh_v_i16m2 (const int16_t *a, size_t vl); vint16m4_t __riscv_th_vlh_v_i16m4 (const int16_t *a, size_t vl); vint16m8_t __riscv_th_vlh_v_i16m8 (const int16_t *a, size_t vl); vint32m1_t __riscv_th_vlh_v_i32m1 (const int32_t *a, size_t vl); vint32m2_t __riscv_th_vlh_v_i32m2 (const int32_t *a, size_t vl); vint32m4_t __riscv_th_vlh_v_i32m4 (const int32_t *a, size_t vl); vint32m8_t __riscv_th_vlh_v_i32m8 (const int32_t *a, size_t vl); vint64m1_t __riscv_th_vlh_v_i64m1 (const int64_t *a, size_t vl); vint64m2_t __riscv_th_vlh_v_i64m2 (const int64_t *a, size_t vl); vint64m4_t __riscv_th_vlh_v_i64m4 (const int64_t *a, size_t vl); vint64m8_t __riscv_th_vlh_v_i64m8 (const int64_t *a, size_t vl); vint32m1_t __riscv_th_vlw_v_i32m1 (const int32_t *a, size_t vl); vint32m2_t __riscv_th_vlw_v_i32m2 (const int32_t *a, size_t vl); vint32m4_t __riscv_th_vlw_v_i32m4 (const int32_t *a, size_t vl); vint32m8_t __riscv_th_vlw_v_i32m8 (const int32_t *a, size_t vl); vint64m1_t __riscv_th_vlw_v_i64m1 (const int64_t *a, size_t vl); vint64m2_t __riscv_th_vlw_v_i64m2 (const int64_t *a, size_t vl); vint64m4_t __riscv_th_vlw_v_i64m4 (const int64_t *a, size_t vl); vint64m8_t __riscv_th_vlw_v_i64m8 (const int64_t *a, size_t vl); With the exisiting framework, I cannot come up with better way to differentiate between vlb/vlw.vlh. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月10日(星期三) 19:09 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. So vlb has not only sew = 8 ? But why do you add intrinsics as follows ? +DEF_RVV_FUNCTION (th_vlb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_ops) Why it is not : DEF_RVV_FUNCTION (th_vlb, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) ? juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-10 19:06 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. The key difference between vlb/vlh/vlw is not output type too. Their difference is the range of datatype, not one specific type. We have dived into the xtheadvector special intrinsics and are sure about that. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月10日(星期三) 19:00 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. instance.op_info->args[i].get_tree_type (instance.type.index) is output type. You can use GDB debug it . juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-10 18:57 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. Hi Juzhe, Perhaps things are not as simple as imagined. The differences between vlb/vlh/vlw is not the same as vle8/vle16/vle32. "8", "16" or "32" in vle8/vle16/vle32 can be appended from "vle" according to input type. But vlb/vlh/vlw is different not in input type. ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月10日(星期三) 18:03 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. I mean change these: +DEF_RVV_FUNCTION (th_vlb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_ops) into a single: +DEF_RVV_FUNCTION (th_vl, th_loadstore_width, full_preds, all_v_scalar_const_ptr_ops) and append "h", "w", or"b" according to TYPE_UNSIGNED and GET_MODE_BITSIZE (GET_MODE_INNER (TYPE_MODE (instance.op_info->args[i].get_tree_type (instance.type.index)))) in th_loadstore_width. It should definitely works, I allow this flexibility in design of the framework. juzhe.zhong@rivai.ai 发件人： joshua 发送时间： 2024-01-10 17:55 收件人： juzhe.zhong@rivai.ai; gcc-patches 抄送： Jim Wilson; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; jinma; cooper.qu 主题： Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. And revise th_loadstore_width, append the name according TYPE_UNSIGNED and GET_MODE_BITSIZE (GET_MODE_INNER (TYPE_MODE (instance.op_info->args[i].get_tree_type (instance.type.index)))) What do you mean by it? I'm a bit confused. Changing i8_v_scalar_const_ptr_ops into all_v_scalar_const_ptr_ops will expand the datatypes that can be used in th_vlb. Can we restrict again in th_loadstore_width? ------------------------------------------------------------------ 发件人：juzhe.zhong@rivai.ai <juzhe.zhong@rivai.ai> 发送时间：2024年1月10日(星期三) 17:35 收件人："cooper.joshua"<cooper.joshua@linux.alibaba.com>; "gcc-patches"<gcc-patches@gcc.gnu.org> 抄 送：Jim Wilson<jim.wilson.gcc@gmail.com>; palmer<palmer@dabbelt.com>; andrew<andrew@sifive.com>; "philipp.tomsich"<philipp.tomsich@vrull.eu>; jeffreyalaw<jeffreyalaw@gmail.com>; "christoph.muellner"<christoph.muellner@vrull.eu>; "cooper.joshua"<cooper.joshua@linux.alibaba.com>; jinma<jinma@linux.alibaba.com>; "cooper.qu"<cooper.qu@linux.alibaba.com> 主 题：Re: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. +DEF_RVV_FUNCTION (th_vlb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_ops) I think we should remove those many data structure you added like: i8_v_scalar_const_ptr_ops Instead, you should use all_v_scalar_const_ptr_ops And revise th_loadstore_width, append the name according TYPE_UNSIGNED and GET_MODE_BITSIZE (GET_MODE_INNER (TYPE_MODE (instance.op_info->args[i].get_tree_type (instance.type.index)))) juzhe.zhong@rivai.ai From: Jun Sha (Joshua) Date: 2024-01-10 17:27 To: gcc-patches CC: jim.wilson.gcc; palmer; andrew; philipp.tomsich; jeffreyalaw; christoph.muellner; juzhe.zhong; Jun Sha (Joshua); Jin Ma; Xianmiao Qu Subject: [PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics. This patch only involves the generation of xtheadvector special load/store instructions and vext instructions. gcc/ChangeLog: * config/riscv/riscv-vector-builtins-bases.cc (class th_loadstore_width): Define new builtin bases. (BASE): Define new builtin bases. * config/riscv/riscv-vector-builtins-bases.h: Define new builtin class. * config/riscv/riscv-vector-builtins-functions.def (vlsegff): Include thead-vector-builtins-functions.def. * config/riscv/riscv-vector-builtins-shapes.cc (struct th_loadstore_width_def): Define new builtin shapes. (struct th_indexed_loadstore_width_def): Define new builtin shapes. (SHAPE): Define new builtin shapes. * config/riscv/riscv-vector-builtins-shapes.h: Define new builtin shapes. * config/riscv/riscv-vector-builtins-types.def (DEF_RVV_I8_OPS): Add datatypes for XTheadVector. (DEF_RVV_I16_OPS): Add datatypes for XTheadVector. (DEF_RVV_I32_OPS): Add datatypes for XTheadVector. (DEF_RVV_U8_OPS): Add datatypes for XTheadVector. (DEF_RVV_U16_OPS): Add datatypes for XTheadVector. (DEF_RVV_U32_OPS): Add datatypes for XTheadVector. (vint8m1_t): Add datatypes for XTheadVector. (vint8m2_t): Likewise. (vint8m4_t): Likewise. (vint8m8_t): Likewise. (vint16m1_t): Likewise. (vint16m2_t): Likewise. (vint16m4_t): Likewise. (vint16m8_t): Likewise. (vint32m1_t): Likewise. (vint32m2_t): Likewise. (vint32m4_t): Likewise. (vint32m8_t): Likewise. (vint64m1_t): Likewise. (vint64m2_t): Likewise. (vint64m4_t): Likewise. (vint64m8_t): Likewise. (vuint8m1_t): Likewise. (vuint8m2_t): Likewise. (vuint8m4_t): Likewise. (vuint8m8_t): Likewise. (vuint16m1_t): Likewise. (vuint16m2_t): Likewise. (vuint16m4_t): Likewise. (vuint16m8_t): Likewise. (vuint32m1_t): Likewise. (vuint32m2_t): Likewise. (vuint32m4_t): Likewise. (vuint32m8_t): Likewise. (vuint64m1_t): Likewise. (vuint64m2_t): Likewise. (vuint64m4_t): Likewise. (vuint64m8_t): Likewise. * config/riscv/riscv-vector-builtins.cc (DEF_RVV_I8_OPS): Add datatypes for XTheadVector. (DEF_RVV_I16_OPS): Add datatypes for XTheadVector. (DEF_RVV_I32_OPS): Add datatypes for XTheadVector. (DEF_RVV_U8_OPS): Add datatypes for XTheadVector. (DEF_RVV_U16_OPS): Add datatypes for XTheadVector. (DEF_RVV_U32_OPS): Add datatypes for XTheadVector. * config/riscv/thead-vector-builtins-functions.def: New file. * config/riscv/thead-vector.md: Add new patterns. gcc/testsuite/ChangeLog: * gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c: New test. * gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c: New test. Co-authored-by: Jin Ma <jinma@linux.alibaba.com> Co-authored-by: Xianmiao Qu <cooper.qu@linux.alibaba.com> Co-authored-by: Christoph Müllner <christoph.muellner@vrull.eu> --- .../riscv/riscv-vector-builtins-bases.cc | 139 ++++++++ .../riscv/riscv-vector-builtins-bases.h | 31 ++ .../riscv/riscv-vector-builtins-shapes.cc | 98 ++++++ .../riscv/riscv-vector-builtins-shapes.h | 3 + .../riscv/riscv-vector-builtins-types.def | 120 +++++++ gcc/config/riscv/riscv-vector-builtins.cc | 311 ++++++++++++++++++ gcc/config/riscv/riscv-vector-builtins.h | 3 + gcc/config/riscv/t-riscv | 1 + .../riscv/thead-vector-builtins-functions.def | 39 +++ gcc/config/riscv/thead-vector.md | 253 ++++++++++++++ .../riscv/rvv/xtheadvector/vlb-vsb.c | 68 ++++ .../riscv/rvv/xtheadvector/vlbu-vsb.c | 68 ++++ .../riscv/rvv/xtheadvector/vlh-vsh.c | 68 ++++ .../riscv/rvv/xtheadvector/vlhu-vsh.c | 68 ++++ .../riscv/rvv/xtheadvector/vlw-vsw.c | 68 ++++ .../riscv/rvv/xtheadvector/vlwu-vsw.c | 68 ++++ 16 files changed, 1406 insertions(+) create mode 100644 gcc/config/riscv/thead-vector-builtins-functions.def create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc index 46f1a1da33e..5f44f31a12b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.cc +++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc @@ -2125,6 +2125,83 @@ public: } }; +/* Implements + * th.vl(b/h/w)[u].v/th.vs(b/h/w)[u].v/th.vls(b/h/w)[u].v/th.vss(b/h/w)[u].v/ + * th.vlx(b/h/w)[u].v/th.vs[u]x(b/h/w).v + * codegen. */ +template<bool STORE_P, lst_type LST_TYPE, int UNSPEC> +class th_loadstore_width : public function_base +{ +public: + bool apply_tail_policy_p () const override { return !STORE_P; } + bool apply_mask_policy_p () const override { return !STORE_P; } + + unsigned int call_properties (const function_instance &) const override + { + if (STORE_P) + return CP_WRITE_MEMORY; + else + return CP_READ_MEMORY; + } + + bool can_be_overloaded_p (enum predication_type_index pred) const override + { + if (STORE_P || LST_TYPE == LST_INDEXED) + return true; + return pred != PRED_TYPE_none; + } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + if (LST_TYPE == LST_INDEXED) + { + if (STORE_P) + return e.use_exact_insn ( + code_for_pred_indexed_store_width (UNSPEC, UNSPEC, + e.vector_mode ())); + else + return e.use_exact_insn ( + code_for_pred_indexed_load_width (UNSPEC, e.vector_mode ())); + } + else if (LST_TYPE == LST_STRIDED) + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_strided_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_strided_load_width (UNSPEC, e.vector_mode ())); + } + else + { + if (STORE_P) + return e.use_contiguous_store_insn ( + code_for_pred_store_width (UNSPEC, e.vector_mode ())); + else + return e.use_contiguous_load_insn ( + code_for_pred_mov_width (UNSPEC, e.vector_mode ())); + } + } +}; + +/* Implements vext.x.v. */ +class th_extract : public function_base +{ +public: + bool apply_vl_p () const override { return false; } + bool apply_tail_policy_p () const override { return false; } + bool apply_mask_policy_p () const override { return false; } + bool use_mask_predication_p () const override { return false; } + bool has_merge_operand_p () const override { return false; } + + rtx expand (function_expander &e) const override + { + gcc_assert (TARGET_XTHEADVECTOR); + return e.use_exact_insn (code_for_pred_th_extract (e.vector_mode ())); + } +}; + /* Below implements are vector crypto */ /* Implements vandn.[vv,vx] */ class vandn : public function_base @@ -2587,6 +2664,37 @@ static CONSTEXPR const seg_indexed_load<UNSPEC_ORDERED> vloxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_UNORDERED> vsuxseg_obj; static CONSTEXPR const seg_indexed_store<UNSPEC_ORDERED> vsoxseg_obj; static CONSTEXPR const vlsegff vlsegff_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLB> th_vlb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLBU> th_vlbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLH> th_vlh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLHU> th_vlhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLW> th_vlw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_UNIT_STRIDE, UNSPEC_TH_VLWU> th_vlwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLB> th_vsb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLH> th_vsh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_UNIT_STRIDE, UNSPEC_TH_VLW> th_vsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSB> th_vlsb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSBU> th_vlsbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSH> th_vlsh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSHU> th_vlshu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSW> th_vlsw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_STRIDED, UNSPEC_TH_VLSWU> th_vlswu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSB> th_vssb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSH> th_vssh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_STRIDED, UNSPEC_TH_VLSW> th_vssw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXB> th_vlxb_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXBU> th_vlxbu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXH> th_vlxh_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXHU> th_vlxhu_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXW> th_vlxw_obj; +static CONSTEXPR const th_loadstore_width<false, LST_INDEXED, UNSPEC_TH_VLXWU> th_vlxwu_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXB> th_vsxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXH> th_vsxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VLXW> th_vsxw_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXB> th_vsuxb_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXH> th_vsuxh_obj; +static CONSTEXPR const th_loadstore_width<true, LST_INDEXED, UNSPEC_TH_VSUXW> th_vsuxw_obj; +static CONSTEXPR const th_extract th_vext_x_v_obj; /* Crypto Vector */ static CONSTEXPR const vandn vandn_obj; @@ -2878,6 +2986,37 @@ BASE (vloxseg) BASE (vsuxseg) BASE (vsoxseg) BASE (vlsegff) +BASE (th_vlb) +BASE (th_vlh) +BASE (th_vlw) +BASE (th_vlbu) +BASE (th_vlhu) +BASE (th_vlwu) +BASE (th_vsb) +BASE (th_vsh) +BASE (th_vsw) +BASE (th_vlsb) +BASE (th_vlsh) +BASE (th_vlsw) +BASE (th_vlsbu) +BASE (th_vlshu) +BASE (th_vlswu) +BASE (th_vssb) +BASE (th_vssh) +BASE (th_vssw) +BASE (th_vlxb) +BASE (th_vlxh) +BASE (th_vlxw) +BASE (th_vlxbu) +BASE (th_vlxhu) +BASE (th_vlxwu) +BASE (th_vsxb) +BASE (th_vsxh) +BASE (th_vsxw) +BASE (th_vsuxb) +BASE (th_vsuxh) +BASE (th_vsuxw) +BASE (th_vext_x_v) /* Crypto vector */ BASE (vandn) BASE (vbrev) diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h index 1122e3801a7..df43adf9a17 100644 --- a/gcc/config/riscv/riscv-vector-builtins-bases.h +++ b/gcc/config/riscv/riscv-vector-builtins-bases.h @@ -299,6 +299,37 @@ extern const function_base *const vloxseg; extern const function_base *const vsuxseg; extern const function_base *const vsoxseg; extern const function_base *const vlsegff; +extern const function_base *const th_vlb; +extern const function_base *const th_vlh; +extern const function_base *const th_vlw; +extern const function_base *const th_vlbu; +extern const function_base *const th_vlhu; +extern const function_base *const th_vlwu; +extern const function_base *const th_vsb; +extern const function_base *const th_vsh; +extern const function_base *const th_vsw; +extern const function_base *const th_vlsb; +extern const function_base *const th_vlsh; +extern const function_base *const th_vlsw; +extern const function_base *const th_vlsbu; +extern const function_base *const th_vlshu; +extern const function_base *const th_vlswu; +extern const function_base *const th_vssb; +extern const function_base *const th_vssh; +extern const function_base *const th_vssw; +extern const function_base *const th_vlxb; +extern const function_base *const th_vlxh; +extern const function_base *const th_vlxw; +extern const function_base *const th_vlxbu; +extern const function_base *const th_vlxhu; +extern const function_base *const th_vlxwu; +extern const function_base *const th_vsxb; +extern const function_base *const th_vsxh; +extern const function_base *const th_vsxw; +extern const function_base *const th_vsuxb; +extern const function_base *const th_vsuxh; +extern const function_base *const th_vsuxw; +extern const function_base *const th_vext_x_v; /* Below function_base are Vectro Crypto*/ extern const function_base *const vandn; extern const function_base *const vbrev; diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.cc b/gcc/config/riscv/riscv-vector-builtins-shapes.cc index 1e4f4d53de6..489a95cf684 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.cc +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.cc @@ -211,6 +211,86 @@ struct indexed_loadstore_def : public function_shape } }; +/* th_loadstore_width_def class. */ +struct th_loadstore_width_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_base_name (instance.base_name); + + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + + +/* th_indexed_loadstore_width_def class. */ +struct th_indexed_loadstore_width_def : public function_shape +{ + void build (function_builder &b, + const function_group_info &group) const override + { + for (unsigned int pred_idx = 0; group.preds[pred_idx] != NUM_PRED_TYPES; + ++pred_idx) + { + for (unsigned int vec_type_idx = 0; + group.ops_infos.types[vec_type_idx].index != NUM_VECTOR_TYPES; + ++vec_type_idx) + { + tree index_type = group.ops_infos.args[1].get_tree_type ( + group.ops_infos.types[vec_type_idx].index); + if (!index_type) + continue; + build_one (b, group, pred_idx, vec_type_idx); + } + } + } + + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + + /* Return nullptr if it can not be overloaded. */ + if (overloaded_p && !instance.base->can_be_overloaded_p (instance.pred)) + return nullptr; + + b.append_base_name (instance.base_name); + /* vop_v --> vop_v_<type>. */ + if (!overloaded_p) + { + /* vop --> vop_v. */ + b.append_name (operand_suffixes[instance.op_info->op]); + /* vop_v --> vop_v_<type>. */ + b.append_name (type_suffixes[instance.type.index].vector); + } + + /* According to rvv-intrinsic-doc, it does not add "_m" suffix + for vop_m C++ overloaded API. */ + if (overloaded_p && instance.pred == PRED_TYPE_m) + return b.finish_name (); + b.append_name (predication_suffixes[instance.pred]); + return b.finish_name (); + } +}; + /* alu_def class. */ struct alu_def : public build_base { @@ -632,6 +712,21 @@ struct reduc_alu_def : public build_base } }; +/* th_extract_def class. */ +struct th_extract_def : public build_base +{ + char *get_name (function_builder &b, const function_instance &instance, + bool overloaded_p) const override + { + b.append_base_name (instance.base_name); + if (overloaded_p) + return b.finish_name (); + b.append_name (type_suffixes[instance.type.index].vector); + b.append_name (type_suffixes[instance.type.index].scalar); + return b.finish_name (); + } +}; + /* scalar_move_def class. */ struct scalar_move_def : public build_base { @@ -1094,6 +1189,8 @@ SHAPE(vsetvl, vsetvl) SHAPE(vsetvl, vsetvlmax) SHAPE(loadstore, loadstore) SHAPE(indexed_loadstore, indexed_loadstore) +SHAPE(th_loadstore_width, th_loadstore_width) +SHAPE(th_indexed_loadstore_width, th_indexed_loadstore_width) SHAPE(alu, alu) SHAPE(alu_frm, alu_frm) SHAPE(widen_alu, widen_alu) @@ -1106,6 +1203,7 @@ SHAPE(move, move) SHAPE(mask_alu, mask_alu) SHAPE(reduc_alu, reduc_alu) SHAPE(reduc_alu_frm, reduc_alu_frm) +SHAPE(th_extract, th_extract) SHAPE(scalar_move, scalar_move) SHAPE(vundefined, vundefined) SHAPE(misc, misc) diff --git a/gcc/config/riscv/riscv-vector-builtins-shapes.h b/gcc/config/riscv/riscv-vector-builtins-shapes.h index ac2a28ce017..a7624d0fabd 100644 --- a/gcc/config/riscv/riscv-vector-builtins-shapes.h +++ b/gcc/config/riscv/riscv-vector-builtins-shapes.h @@ -28,6 +28,8 @@ extern const function_shape *const vsetvl; extern const function_shape *const vsetvlmax; extern const function_shape *const loadstore; extern const function_shape *const indexed_loadstore; +extern const function_shape *const th_loadstore_width; +extern const function_shape *const th_indexed_loadstore_width; extern const function_shape *const alu; extern const function_shape *const alu_frm; extern const function_shape *const widen_alu; @@ -41,6 +43,7 @@ extern const function_shape *const mask_alu; extern const function_shape *const reduc_alu; extern const function_shape *const reduc_alu_frm; extern const function_shape *const scalar_move; +extern const function_shape *const th_extract; extern const function_shape *const vundefined; extern const function_shape *const misc; extern const function_shape *const vset; diff --git a/gcc/config/riscv/riscv-vector-builtins-types.def b/gcc/config/riscv/riscv-vector-builtins-types.def index 61019a56844..abfeb4fcd9b 100644 --- a/gcc/config/riscv/riscv-vector-builtins-types.def +++ b/gcc/config/riscv/riscv-vector-builtins-types.def @@ -24,12 +24,48 @@ along with GCC; see the file COPYING3. If not see #define DEF_RVV_I_OPS(TYPE, REQUIRE) #endif +/* Use "DEF_RVV_I8_OPS" macro include some signed integer (i8/i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I8_OPS +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_I16_OPS" macro include some signed integer (i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I16_OPS +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_I32_OPS" macro include some signed integer (i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_I32_OPS +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) +#endif + /* Use "DEF_RVV_U_OPS" macro include all unsigned integer which will be iterated and registered as intrinsic functions. */ #ifndef DEF_RVV_U_OPS #define DEF_RVV_U_OPS(TYPE, REQUIRE) #endif +/* Use "DEF_RVV_U8_OPS" macro include some unsigned integer (i8/i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U8_OPS +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_U16_OPS" macro include some unsigned integer (i16/i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U16_OPS +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) +#endif + +/* Use "DEF_RVV_U32_OPS" macro include some unsigned integer (i32/i64) + which will be iterated and registered as intrinsic functions. */ +#ifndef DEF_RVV_U32_OPS +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) +#endif + /* Use "DEF_RVV_F_OPS" macro include all floating-point which will be iterated and registered as intrinsic functions. */ #ifndef DEF_RVV_F_OPS @@ -374,6 +410,45 @@ DEF_RVV_I_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) DEF_RVV_I_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) DEF_RVV_I_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint8m1_t, 0) +DEF_RVV_I8_OPS (vint8m2_t, 0) +DEF_RVV_I8_OPS (vint8m4_t, 0) +DEF_RVV_I8_OPS (vint8m8_t, 0) +DEF_RVV_I8_OPS (vint16m1_t, 0) +DEF_RVV_I8_OPS (vint16m2_t, 0) +DEF_RVV_I8_OPS (vint16m4_t, 0) +DEF_RVV_I8_OPS (vint16m8_t, 0) +DEF_RVV_I8_OPS (vint32m1_t, 0) +DEF_RVV_I8_OPS (vint32m2_t, 0) +DEF_RVV_I8_OPS (vint32m4_t, 0) +DEF_RVV_I8_OPS (vint32m8_t, 0) +DEF_RVV_I8_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I8_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_I16_OPS (vint16m1_t, 0) +DEF_RVV_I16_OPS (vint16m2_t, 0) +DEF_RVV_I16_OPS (vint16m4_t, 0) +DEF_RVV_I16_OPS (vint16m8_t, 0) +DEF_RVV_I16_OPS (vint32m1_t, 0) +DEF_RVV_I16_OPS (vint32m2_t, 0) +DEF_RVV_I16_OPS (vint32m4_t, 0) +DEF_RVV_I16_OPS (vint32m8_t, 0) +DEF_RVV_I16_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I16_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_I32_OPS (vint32m1_t, 0) +DEF_RVV_I32_OPS (vint32m2_t, 0) +DEF_RVV_I32_OPS (vint32m4_t, 0) +DEF_RVV_I32_OPS (vint32m8_t, 0) +DEF_RVV_I32_OPS (vint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_I32_OPS (vint64m8_t, RVV_REQUIRE_ELEN_64) + DEF_RVV_U_OPS (vuint8mf8_t, RVV_REQUIRE_MIN_VLEN_64) DEF_RVV_U_OPS (vuint8mf4_t, 0) DEF_RVV_U_OPS (vuint8mf2_t, 0) @@ -397,6 +472,45 @@ DEF_RVV_U_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) DEF_RVV_U_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) DEF_RVV_U_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint8m1_t, 0) +DEF_RVV_U8_OPS (vuint8m2_t, 0) +DEF_RVV_U8_OPS (vuint8m4_t, 0) +DEF_RVV_U8_OPS (vuint8m8_t, 0) +DEF_RVV_U8_OPS (vuint16m1_t, 0) +DEF_RVV_U8_OPS (vuint16m2_t, 0) +DEF_RVV_U8_OPS (vuint16m4_t, 0) +DEF_RVV_U8_OPS (vuint16m8_t, 0) +DEF_RVV_U8_OPS (vuint32m1_t, 0) +DEF_RVV_U8_OPS (vuint32m2_t, 0) +DEF_RVV_U8_OPS (vuint32m4_t, 0) +DEF_RVV_U8_OPS (vuint32m8_t, 0) +DEF_RVV_U8_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U8_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_U16_OPS (vuint16m1_t, 0) +DEF_RVV_U16_OPS (vuint16m2_t, 0) +DEF_RVV_U16_OPS (vuint16m4_t, 0) +DEF_RVV_U16_OPS (vuint16m8_t, 0) +DEF_RVV_U16_OPS (vuint32m1_t, 0) +DEF_RVV_U16_OPS (vuint32m2_t, 0) +DEF_RVV_U16_OPS (vuint32m4_t, 0) +DEF_RVV_U16_OPS (vuint32m8_t, 0) +DEF_RVV_U16_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U16_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + +DEF_RVV_U32_OPS (vuint32m1_t, 0) +DEF_RVV_U32_OPS (vuint32m2_t, 0) +DEF_RVV_U32_OPS (vuint32m4_t, 0) +DEF_RVV_U32_OPS (vuint32m8_t, 0) +DEF_RVV_U32_OPS (vuint64m1_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m2_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) +DEF_RVV_U32_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) + DEF_RVV_F_OPS (vfloat16mf4_t, RVV_REQUIRE_ELEN_FP_16 | RVV_REQUIRE_MIN_VLEN_64) DEF_RVV_F_OPS (vfloat16mf2_t, RVV_REQUIRE_ELEN_FP_16) DEF_RVV_F_OPS (vfloat16m1_t, RVV_REQUIRE_ELEN_FP_16) @@ -1379,7 +1493,13 @@ DEF_RVV_CRYPTO_SEW64_OPS (vuint64m4_t, RVV_REQUIRE_ELEN_64) DEF_RVV_CRYPTO_SEW64_OPS (vuint64m8_t, RVV_REQUIRE_ELEN_64) #undef DEF_RVV_I_OPS +#undef DEF_RVV_I8_OPS +#undef DEF_RVV_I16_OPS +#undef DEF_RVV_I32_OPS #undef DEF_RVV_U_OPS +#undef DEF_RVV_U8_OPS +#undef DEF_RVV_U16_OPS +#undef DEF_RVV_U32_OPS #undef DEF_RVV_F_OPS #undef DEF_RVV_B_OPS #undef DEF_RVV_WEXTI_OPS diff --git a/gcc/config/riscv/riscv-vector-builtins.cc b/gcc/config/riscv/riscv-vector-builtins.cc index 25e0b6e56de..f429f12dc18 100644 --- a/gcc/config/riscv/riscv-vector-builtins.cc +++ b/gcc/config/riscv/riscv-vector-builtins.cc @@ -246,6 +246,63 @@ static const rvv_type_info iu_ops[] = { #include "riscv-vector-builtins-types.def" {NUM_VECTOR_TYPES, 0}}; +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i8_ops[] = { +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i16_ops[] = { +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info i32_ops[] = { +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u8_ops[] = { +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u16_ops[] = { +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info u32_ops[] = { +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu8_ops[] = { +#define DEF_RVV_I8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U8_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu16_ops[] = { +#define DEF_RVV_I16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U16_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + +/* A list of all integer will be registered for intrinsic functions. */ +static const rvv_type_info iu32_ops[] = { +#define DEF_RVV_I32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#define DEF_RVV_U32_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, +#include "riscv-vector-builtins-types.def" + {NUM_VECTOR_TYPES, 0}}; + /* A list of all types will be registered for intrinsic functions. */ static const rvv_type_info all_ops[] = { #define DEF_RVV_I_OPS(TYPE, REQUIRE) {VECTOR_TYPE_##TYPE, REQUIRE}, @@ -934,6 +991,32 @@ static CONSTEXPR const rvv_arg_type_info ext_vcreate_args[] = {rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; +/* A list of args for vector_type func (const scalar_type *, size_t) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info_end}; + +/* A list of args for vector_type func (const scalar_type *, eew8_index_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_const_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_const_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, eew8_index_type, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_index_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_unsigned_vector), + rvv_arg_type_info (RVV_BASE_vector), rvv_arg_type_info_end}; + +/* A list of args for void func (scalar_type *, size_t, vector_type) + * function. */ +static CONSTEXPR const rvv_arg_type_info scalar_ptr_size_args[] + = {rvv_arg_type_info (RVV_BASE_scalar_ptr), + rvv_arg_type_info (RVV_BASE_size), rvv_arg_type_info (RVV_BASE_vector), + rvv_arg_type_info_end}; + /* A list of none preds that will be registered for intrinsic functions. */ static CONSTEXPR const predication_type_index none_preds[] = {PRED_TYPE_none, NUM_PRED_TYPES}; @@ -1455,6 +1538,14 @@ static CONSTEXPR const rvv_op_info iu_shift_vvv_ops rvv_arg_type_info (RVV_BASE_vector), /* Return type */ shift_vv_args /* Args */}; +/* A static operand information for scalar_type func (vector_type, size_t) + * function registration. */ +static CONSTEXPR const rvv_op_info iu_x_s_u_ops + = {iu_ops, /* Types */ + OP_TYPE_vx, /* Suffix */ + rvv_arg_type_info (RVV_BASE_scalar), /* Return type */ + v_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type, size_t) * function registration. */ static CONSTEXPR const rvv_op_info iu_shift_vvx_ops @@ -2638,6 +2729,222 @@ static CONSTEXPR const rvv_op_info all_v_vcreate_lmul4_x2_ops rvv_arg_type_info (RVV_BASE_vlmul_ext_x2), /* Return type */ ext_vcreate_args /* Args */}; +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *) + * function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_size_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_size_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_size_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_size_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_size_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * size_t) function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_size_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_size_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i8_v_scalar_const_ptr_index_ops + = {i8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u8_v_scalar_const_ptr_index_ops + = {u8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i16_v_scalar_const_ptr_index_ops + = {i16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u16_v_scalar_const_ptr_index_ops + = {u16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info i32_v_scalar_const_ptr_index_ops + = {i32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for vector_type func (const scalar_type *, + * eew8_index_type) function registration. */ +static CONSTEXPR const rvv_op_info u32_v_scalar_const_ptr_index_ops + = {u32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_vector), /* Return type */ + scalar_const_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew8_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_index_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew16_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_index_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, eew32_index_type, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_index_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_index_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type, + * function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type) + * function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, vector_type) + * function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu8_v_scalar_ptr_size_ops + = {iu8_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu16_v_scalar_ptr_size_ops + = {iu16_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + +/* A static operand information for void func (scalar_type *, size_t, + * vector_type) function registration. */ +static CONSTEXPR const rvv_op_info iu32_v_scalar_ptr_size_ops + = {iu32_ops, /* Types */ + OP_TYPE_v, /* Suffix */ + rvv_arg_type_info (RVV_BASE_void), /* Return type */ + scalar_ptr_size_args /* Args */}; + /* A static operand information for vector_type func (vector_type). Some ins just supports SEW=32, such as crypto vectol Zvkg extension. * function registration. */ @@ -2816,6 +3123,10 @@ static function_group_info function_groups[] = { #define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, #include "riscv-vector-builtins-functions.def" +#undef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) \ + {#NAME, &bases::NAME, &shapes::SHAPE, PREDS, OPS_INFO, REQUIRED_EXTENSIONS}, +#include "thead-vector-builtins-functions.def" }; /* The RVV types, with their built-in diff --git a/gcc/config/riscv/riscv-vector-builtins.h b/gcc/config/riscv/riscv-vector-builtins.h index 54c8824ff92..a8ee39a3cb2 100644 --- a/gcc/config/riscv/riscv-vector-builtins.h +++ b/gcc/config/riscv/riscv-vector-builtins.h @@ -114,6 +114,7 @@ static const unsigned int CP_WRITE_CSR = 1U << 5; enum required_ext { VECTOR_EXT, /* Vector extension */ + XTHEADVECTOR_EXT, /* XTheadVector extension */ ZVBB_EXT, /* Cryto vector Zvbb sub-ext */ ZVBB_OR_ZVKB_EXT, /* Cryto vector Zvbb or zvkb sub-ext */ ZVBC_EXT, /* Crypto vector Zvbc sub-ext */ @@ -234,6 +235,8 @@ struct function_group_info { case VECTOR_EXT: return TARGET_VECTOR; + case XTHEADVECTOR_EXT: + return TARGET_XTHEADVECTOR; case ZVBB_EXT: return TARGET_ZVBB; case ZVBB_OR_ZVKB_EXT: diff --git a/gcc/config/riscv/t-riscv b/gcc/config/riscv/t-riscv index 32de6b851c1..38494320d8b 100644 --- a/gcc/config/riscv/t-riscv +++ b/gcc/config/riscv/t-riscv @@ -1,6 +1,7 @@ RISCV_BUILTINS_H = $(srcdir)/config/riscv/riscv-vector-builtins.h \ $(srcdir)/config/riscv/riscv-vector-builtins.def \ $(srcdir)/config/riscv/riscv-vector-builtins-functions.def \ + $(srcdir)/config/riscv/thead-vector-builtins-functions.def \ riscv-vector-type-indexer.gen.def riscv-builtins.o: $(srcdir)/config/riscv/riscv-builtins.cc $(CONFIG_H) \ diff --git a/gcc/config/riscv/thead-vector-builtins-functions.def b/gcc/config/riscv/thead-vector-builtins-functions.def new file mode 100644 index 00000000000..667820d4c3e --- /dev/null +++ b/gcc/config/riscv/thead-vector-builtins-functions.def @@ -0,0 +1,39 @@ +#ifndef DEF_RVV_FUNCTION +#define DEF_RVV_FUNCTION(NAME, SHAPE, PREDS, OPS_INFO) +#endif + +#define REQUIRED_EXTENSIONS XTHEADVECTOR_EXT +DEF_RVV_FUNCTION (th_vlb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlbu, th_loadstore_width, full_preds, u8_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlhu, th_loadstore_width, full_preds, u16_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vlwu, th_loadstore_width, full_preds, u32_v_scalar_const_ptr_ops) +DEF_RVV_FUNCTION (th_vsb, th_loadstore_width, none_m_preds, iu8_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vsh, th_loadstore_width, none_m_preds, iu16_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vsw, th_loadstore_width, none_m_preds, iu32_v_scalar_ptr_ops) +DEF_RVV_FUNCTION (th_vlsb, th_loadstore_width, full_preds, i8_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsh, th_loadstore_width, full_preds, i16_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsw, th_loadstore_width, full_preds, i32_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlsbu, th_loadstore_width, full_preds, u8_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlshu, th_loadstore_width, full_preds, u16_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlswu, th_loadstore_width, full_preds, u32_v_scalar_const_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssb, th_loadstore_width, none_m_preds, iu8_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssh, th_loadstore_width, none_m_preds, iu16_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vssw, th_loadstore_width, none_m_preds, iu32_v_scalar_ptr_size_ops) +DEF_RVV_FUNCTION (th_vlxb, th_indexed_loadstore_width, full_preds, i8_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxh, th_indexed_loadstore_width, full_preds, i16_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxw, th_indexed_loadstore_width, full_preds, i32_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxbu, th_indexed_loadstore_width, full_preds, u8_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxhu, th_indexed_loadstore_width, full_preds, u16_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vlxwu, th_indexed_loadstore_width, full_preds, u32_v_scalar_const_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxb, th_indexed_loadstore_width, none_m_preds, iu8_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxh, th_indexed_loadstore_width, none_m_preds, iu16_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsxw, th_indexed_loadstore_width, none_m_preds, iu32_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxb, th_indexed_loadstore_width, none_m_preds, iu8_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxh, th_indexed_loadstore_width, none_m_preds, iu16_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vsuxw, th_indexed_loadstore_width, none_m_preds, iu32_v_scalar_ptr_index_ops) +DEF_RVV_FUNCTION (th_vext_x_v, th_extract, none_preds, iu_x_s_u_ops) +#undef REQUIRED_EXTENSIONS + +#undef DEF_RVV_FUNCTION diff --git a/gcc/config/riscv/thead-vector.md b/gcc/config/riscv/thead-vector.md index 696b815252d..0f3700d9269 100644 --- a/gcc/config/riscv/thead-vector.md +++ b/gcc/config/riscv/thead-vector.md @@ -1,7 +1,95 @@ (define_c_enum "unspec" [ + UNSPEC_TH_VLB + UNSPEC_TH_VLBU + UNSPEC_TH_VLH + UNSPEC_TH_VLHU + UNSPEC_TH_VLW + UNSPEC_TH_VLWU + + UNSPEC_TH_VLSB + UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH + UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW + UNSPEC_TH_VLSWU + + UNSPEC_TH_VLXB + UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH + UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW + UNSPEC_TH_VLXWU + + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW + UNSPEC_TH_VWLDST ]) +(define_int_iterator UNSPEC_TH_VLMEM_OP [ + UNSPEC_TH_VLB UNSPEC_TH_VLBU + UNSPEC_TH_VLH UNSPEC_TH_VLHU + UNSPEC_TH_VLW UNSPEC_TH_VLWU +]) + +(define_int_iterator UNSPEC_TH_VLSMEM_OP [ + UNSPEC_TH_VLSB UNSPEC_TH_VLSBU + UNSPEC_TH_VLSH UNSPEC_TH_VLSHU + UNSPEC_TH_VLSW UNSPEC_TH_VLSWU +]) + +(define_int_iterator UNSPEC_TH_VLXMEM_OP [ + UNSPEC_TH_VLXB UNSPEC_TH_VLXBU + UNSPEC_TH_VLXH UNSPEC_TH_VLXHU + UNSPEC_TH_VLXW UNSPEC_TH_VLXWU +]) + +(define_int_attr vlmem_op_attr [ + (UNSPEC_TH_VLB "b") (UNSPEC_TH_VLBU "bu") + (UNSPEC_TH_VLH "h") (UNSPEC_TH_VLHU "hu") + (UNSPEC_TH_VLW "w") (UNSPEC_TH_VLWU "wu") + (UNSPEC_TH_VLSB "b") (UNSPEC_TH_VLSBU "bu") + (UNSPEC_TH_VLSH "h") (UNSPEC_TH_VLSHU "hu") + (UNSPEC_TH_VLSW "w") (UNSPEC_TH_VLSWU "wu") + (UNSPEC_TH_VLXB "b") (UNSPEC_TH_VLXBU "bu") + (UNSPEC_TH_VLXH "h") (UNSPEC_TH_VLXHU "hu") + (UNSPEC_TH_VLXW "w") (UNSPEC_TH_VLXWU "wu") + (UNSPEC_TH_VSUXB "b") + (UNSPEC_TH_VSUXH "h") + (UNSPEC_TH_VSUXW "w") +]) + +(define_int_attr vlmem_order_attr [ + (UNSPEC_TH_VLXB "") + (UNSPEC_TH_VLXH "") + (UNSPEC_TH_VLXW "") + (UNSPEC_TH_VSUXB "u") + (UNSPEC_TH_VSUXH "u") + (UNSPEC_TH_VSUXW "u") +]) + +(define_int_iterator UNSPEC_TH_VSMEM_OP [ + UNSPEC_TH_VLB + UNSPEC_TH_VLH + UNSPEC_TH_VLW +]) + +(define_int_iterator UNSPEC_TH_VSSMEM_OP [ + UNSPEC_TH_VLSB + UNSPEC_TH_VLSH + UNSPEC_TH_VLSW +]) + +(define_int_iterator UNSPEC_TH_VSXMEM_OP [ + UNSPEC_TH_VLXB + UNSPEC_TH_VLXH + UNSPEC_TH_VLXW + UNSPEC_TH_VSUXB + UNSPEC_TH_VSUXH + UNSPEC_TH_VSUXW +]) + (define_mode_iterator V_VLS_VT [V VLS VT]) (define_mode_iterator V_VB_VLS_VT [V VB VLS VT]) @@ -100,3 +188,168 @@ } [(set_attr "type" "vldm,vstm,vmalu,vmalu,vmalu") (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand") + (match_operand 4 "vector_length_operand") + (match_operand 5 "const_int_operand") + (match_operand 6 "const_int_operand") + (match_operand 7 "const_int_operand") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "vector_move_operand") + (match_operand:V_VLS 2 "vector_merge_operand")))] + "TARGET_XTHEADVECTOR" + {}) + +(define_insn_and_split "*pred_mov_width<vlmem_op_attr><mode>" + [(set (match_operand:V_VLS 0 "nonimmediate_operand" "=vr, vr, vd, m, vr, vr") + (if_then_else:V_VLS + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm, vmWc1, Wc1, Wc1") + (match_operand 4 "vector_length_operand" " rK, rK, rK, rK, rK, rK") + (match_operand 5 "const_int_operand" " i, i, i, i, i, i") + (match_operand 6 "const_int_operand" " i, i, i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLMEM_OP) + (match_operand:V_VLS 3 "reg_or_mem_operand" " m, m, m, vr, vr, vr") + (match_operand:V_VLS 2 "vector_merge_operand" " 0, vu, vu, vu, vu, 0")))] + "(TARGET_XTHEADVECTOR + && (register_operand (operands[0], <MODE>mode) + || register_operand (operands[3], <MODE>mode)))" + "@ + vl<vlmem_op_attr>.v\t%0,%3%p1 + vl<vlmem_op_attr>.v\t%0,%3 + vl<vlmem_op_attr>.v\t%0,%3,%1.t + vs<vlmem_op_attr>.v\t%3,%0%p1 + vmv.v.v\t%0,%3 + vmv.v.v\t%0,%3" + "&& register_operand (operands[0], <MODE>mode) + && register_operand (operands[3], <MODE>mode) + && satisfies_constraint_vu (operands[2]) + && INTVAL (operands[7]) == riscv_vector::VLMAX" + [(set (match_dup 0) (match_dup 3))] + "" + [(set_attr "type" "vlde,vlde,vlde,vste,vimov,vimov") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 3 "vector_length_operand" " rK") + (match_operand 4 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSMEM_OP) + (match_operand:VI 2 "register_operand" " vr") + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_op_attr>.v\t%2,%0%p1" + [(set_attr "type" "vste") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 4)) + (set_attr "vl_op_idx" "3")]) + +(define_insn "@pred_strided_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vr, vr, vd") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1, Wc1, vm") + (match_operand 5 "vector_length_operand" " rK, rK, rK") + (match_operand 6 "const_int_operand" " i, i, i") + (match_operand 7 "const_int_operand" " i, i, i") + (match_operand 8 "const_int_operand" " i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLSMEM_OP) + (unspec:VI + [(match_operand:VI 3 "memory_operand" " m, m, m") + (match_operand 4 "pmode_reg_or_0_operand" " rJ, rJ, rJ")] UNSPEC_TH_VLSMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " 0, vu, vu")))] + "TARGET_XTHEADVECTOR" + "vls<vlmem_op_attr>.v\t%0,%3,%z4%p1" + [(set_attr "type" "vlds") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_strided_store_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "memory_operand" "+m") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSSMEM_OP) + (unspec:VI + [(match_operand 2 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSSMEM_OP) + (match_dup 0)))] + "TARGET_XTHEADVECTOR" + "vss<vlmem_op_attr>.v\t%3,%0,%z2%p1" + [(set_attr "type" "vsts") + (set_attr "mode" "<MODE>") + (set (attr "avl_type_idx") (const_int 5))]) + +(define_insn "@pred_indexed_load_width<vlmem_op_attr><mode>" + [(set (match_operand:VI 0 "register_operand" "=vd, vr,vd, vr") + (if_then_else:VI + (unspec:<VM> + [(match_operand:<VM> 1 "vector_mask_operand" " vm,Wc1,vm,Wc1") + (match_operand 5 "vector_length_operand" " rK, rK,rK, rK") + (match_operand 6 "const_int_operand" " i, i, i, i") + (match_operand 7 "const_int_operand" " i, i, i, i") + (match_operand 8 "const_int_operand" " i, i, i, i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VLXMEM_OP) + (unspec:VI + [(match_operand 3 "pmode_reg_or_0_operand" " rJ, rJ,rJ, rJ") + (mem:BLK (scratch)) + (match_operand:VI 4 "register_operand" " vr, vr,vr, vr")] UNSPEC_TH_VLXMEM_OP) + (match_operand:VI 2 "vector_merge_operand" " vu, vu, 0, 0")))] + "TARGET_XTHEADVECTOR" + "vlx<vlmem_op_attr>.v\t%0,(%z3),%4%p1" + [(set_attr "type" "vldux") + (set_attr "mode" "<MODE>")]) + +(define_insn "@pred_indexed_<vlmem_order_attr>store_width<vlmem_op_attr><mode>" + [(set (mem:BLK (scratch)) + (unspec:BLK + [(unspec:<VM> + [(match_operand:<VM> 0 "vector_mask_operand" "vmWc1") + (match_operand 4 "vector_length_operand" " rK") + (match_operand 5 "const_int_operand" " i") + (reg:SI VL_REGNUM) + (reg:SI VTYPE_REGNUM)] UNSPEC_TH_VSXMEM_OP) + (match_operand 1 "pmode_reg_or_0_operand" " rJ") + (match_operand:VI 2 "register_operand" " vr") + (match_operand:VI 3 "register_operand" " vr")] UNSPEC_TH_VSXMEM_OP))] + "TARGET_XTHEADVECTOR" + "vs<vlmem_order_attr>x<vlmem_op_attr>.v\t%3,(%z1),%2%p0" + [(set_attr "type" "vstux") + (set_attr "mode" "<MODE>")]) + +(define_expand "@pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" +{}) + +(define_insn "*pred_th_extract<mode>" + [(set (match_operand:<VEL> 0 "register_operand" "=r") + (unspec:<VEL> + [(vec_select:<VEL> + (match_operand:V_VLSI 1 "register_operand" "vr") + (parallel [(match_operand:DI 2 "register_operand" "r")])) + (reg:SI VTYPE_REGNUM)] UNSPEC_VPREDICATE))] + "TARGET_XTHEADVECTOR" + "vext.x.v\t%0,%1,%2" + [(set_attr "type" "vimovvx") + (set_attr "mode" "<MODE>")]) diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c new file mode 100644 index 00000000000..4e192bbf025 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlb-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out) +{ + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tu (v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_m (mask, v3, v3, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlb\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlb.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vv\tv[0-9]+,\s*v[0-9]+,\s*v[0-9]+ +** th.vadd\.vv\tv[1-9][0-9]?,\s*v[0-9]+,\s*v[0-9]+,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlb_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlb_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vv_i32m1 (v2, v2, 4); + vint32m1_t v4 = __riscv_vadd_vv_i32m1_tumu (mask, v3, v2, v2, 4); + __riscv_th_vsb_v_i32m1 (out, v4, 4); +} diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c new file mode 100644 index 00000000000..1538afec68e --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlbu-vsb.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsb\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlbu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlbu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsb.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlbu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlbu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsb_v_u32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c new file mode 100644 index 00000000000..bf4924a1d76 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlh-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlh\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlh.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlh_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlh_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, -16, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_i32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c new file mode 100644 index 00000000000..8c451845175 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlhu-vsh.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsh\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlhu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlhu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsh.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlhu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlhu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsh_v_u32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c new file mode 100644 index 00000000000..0f5b09684a5 --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlw-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, int32_t x) +{ + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tu (v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tu (v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_m (mask, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_m (mask, v3, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlw\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlw.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vx\tv[0-9]+,\s*v[0-9]+,\s*[a-x0-9]+ +** th.vadd\.vx\tv[1-9][0-9]?,\s*v[0-9]+,\s*[a-x0-9]+,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, int32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vint32m1_t v = __riscv_th_vlw_v_i32m1 (in, 4); + vint32m1_t v2 = __riscv_th_vlw_v_i32m1_tumu (mask, v, in, 4); + vint32m1_t v3 = __riscv_vadd_vx_i32m1 (v2, x, 4); + vint32m1_t v4 = __riscv_vadd_vx_i32m1_tumu (mask, v3, v2, x, 4); + __riscv_th_vsw_v_i32m1 (out, v4, 4); +} \ No newline at end of file diff --git a/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c new file mode 100644 index 00000000000..aaa75be023d --- /dev/null +++ b/gcc/testsuite/gcc.target/riscv/rvv/xtheadvector/vlwu-vsw.c @@ -0,0 +1,68 @@ +/* { dg-do compile } */ +/* { dg-options "-march=rv32gcxtheadvector -mabi=ilp32d -O3" } */ +/* { dg-final { check-function-bodies "**" "" } } */ +#include "riscv_vector.h" + +/* +** f1: +** th.vsetivli\tzero,4,e32,m1,tu,ma +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vsw\.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f1 (void * in, void *out, uint32_t x) +{ + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tu (v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tu (v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f2: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,ta,ma +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f2 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_m (mask, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_m (mask, v3, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} + +/* +** f3: +** th.vsetvli\t[a-x0-9]+,zero,e8,mf4,ta,ma +** th.vlm.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vsetivli\tzero,4,e32,m1,tu,mu +** th.vlwu\.v\tv[0-9]+,0\([a-x0-9]+\) +** th.vlwu.v\tv[0-9]+,0\([a-x0-9]+\),v0.t +** th.vadd\.vi\tv[0-9]+,\s*v[0-9]+,\s*-16 +** th.vadd\.vi\tv[1-9][0-9]?,\s*v[0-9]+,\s*-16,\s*v0.t +** th.vsw.v\tv[0-9]+,0\([a-x0-9]+\) +** ret +*/ +void f3 (void * in, void *out, uint32_t x) +{ + vbool32_t mask = *(vbool32_t*)in; + asm volatile ("":::"memory"); + vuint32m1_t v = __riscv_th_vlwu_v_u32m1 (in, 4); + vuint32m1_t v2 = __riscv_th_vlwu_v_u32m1_tumu (mask, v, in, 4); + vuint32m1_t v3 = __riscv_vadd_vx_u32m1 (v2, -16, 4); + vuint32m1_t v4 = __riscv_vadd_vx_u32m1_tumu (mask, v3, v2, -16, 4); + __riscv_th_vsw_v_u32m1 (out, v4, 4); +} \ No newline at end of file -- 2.17.1 ^ permalink raw reply [flat|nested] 133+ messages in thread
![[-- Attachment #2: temp4cj.png --] [-- Type: application/octet-stream, Size: 500533 bytes --]](../../5b999f68-c6bc-4333-8263-79defeab44b7.cooper.joshua@linux.alibaba.com/2-temp4cj.png){kind=link}
![[-- Attachment #3: temp4cj.png --] [-- Type: application/octet-stream, Size: 676546 bytes --]](../../5b999f68-c6bc-4333-8263-79defeab44b7.cooper.joshua@linux.alibaba.com/3-temp4cj.png){kind=link}
end of thread, other threads:[~2024-01-19 20:03 UTC | newest]
Thread overview: 133+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-12-29 4:03 [PATCH v4] RISC-V: Support XTheadVector extension Jun Sha (Joshua)
2023-12-29 4:05 ` [PATCH v4] RISC-V: Refactor riscv-vector-builtins-bases.cc Jun Sha (Joshua)
2023-12-29 4:06 ` [PATCH v4] RISC-V: Change csr_operand into Jun Sha (Joshua)
2023-12-29 4:10 ` [PATCH v4] RISC-V: Change csr_operand into vector_length_operand for vsetvl patterns Jun Sha (Joshua)
2024-01-02 1:35 ` juzhe.zhong
2024-01-02 19:50 ` Christoph Müllner
2024-01-02 1:35 ` [PATCH v4] RISC-V: Change csr_operand into juzhe.zhong
2023-12-29 4:13 ` [PATCH v4] RISC-V: Introduce XTheadVector as a subset of V1.0.0 Jun Sha (Joshua)
2024-01-04 9:28 ` Jun Sha (Joshua)
2023-12-29 4:19 ` [PATCH v4] RISC-V: Adds the prefix "th." for the instructions of XTheadVector Jun Sha (Joshua)
2023-12-31 17:43 ` Jeff Law
2024-01-01 22:57 ` 钟居哲
2024-01-03 2:54 ` Andrew Pinski
2024-01-03 3:06 ` juzhe.zhong
2024-01-03 3:19 ` Andrew Pinski
2024-01-03 3:25 ` juzhe.zhong
2024-01-03 3:32 ` Andrew Pinski
2024-01-03 3:32 ` Kito Cheng
2024-01-04 9:15 ` Re：Re: " joshua
2024-01-04 9:18 ` juzhe.zhong
2024-01-04 10:04 ` Christoph Müllner
2024-01-08 2:11 ` Re：Re: " joshua
2024-01-08 3:06 ` Kito Cheng
2024-01-08 3:17 ` Re：Re: " joshua
2024-01-08 3:40 ` Kito Cheng
2024-01-10 3:01 ` Re：Re: " joshua
2024-01-03 6:08 ` Jun Sha (Joshua)
2024-01-08 23:04 ` 钟居哲
2024-01-09 17:49 ` Jeff Law
2024-01-09 22:35 ` 钟居哲
2023-12-29 4:21 ` [PATCH v4] RISC-V: Handle differences between XTheadvector and Vector Jun Sha (Joshua)
2024-01-02 2:00 ` juzhe.zhong
2024-01-02 3:03 ` Re：[PATCH " joshua
2024-01-02 3:10 ` juzhe.zhong
2024-01-02 3:40 ` Re：Re：[PATCH " joshua
2024-01-02 3:41 ` Re：[PATCH " juzhe.zhong
2024-01-02 9:48 ` joshua
2024-01-02 9:52 ` juzhe.zhong
2024-01-02 12:39 ` [PATCH " Jun Sha (Joshua)
2024-01-03 3:11 ` Kito Cheng
2024-01-03 6:15 ` Jun Sha (Joshua)
2024-01-04 2:29 ` Jun Sha (Joshua)
2024-01-09 3:18 ` [PATCH v5] " Jun Sha (Joshua)
2024-01-09 22:33 ` 钟居哲
2024-01-10 2:22 ` Jun Sha (Joshua)
2024-01-10 2:34 ` juzhe.zhong
2024-01-10 2:57 ` Re：[PATCH " joshua
2024-01-10 3:02 ` juzhe.zhong
2024-01-10 6:54 ` juzhe.zhong
2024-01-10 7:01 ` juzhe.zhong
2024-01-10 7:16 ` Re：Re：[PATCH " joshua
2024-01-10 7:17 ` Re：[PATCH " juzhe.zhong
2024-01-10 7:26 ` 回复：Re：[PATCH " joshua
2024-01-10 7:31 ` 回复：[PATCH " juzhe.zhong
2024-01-10 7:28 ` Re：Re：[PATCH " joshua
2024-01-11 11:03 ` [PATCH " Jun Sha (Joshua)
2024-01-08 23:08 ` [PATCH v4] " 钟居哲
2024-01-09 2:12 ` Re：[PATCH " joshua
2024-01-09 3:23 ` joshua
2023-12-29 4:21 ` [PATCH v4 6/6] RISC-V: Add support for xtheadvector-specific intrinsics Jun Sha (Joshua)
2024-01-04 2:34 ` [PATCH v4] " Jun Sha (Joshua)
2024-01-10 9:27 ` [PATCH v5] " Jun Sha (Joshua)
2024-01-10 9:35 ` juzhe.zhong
2024-01-10 9:55 ` Re：[PATCH " joshua
2024-01-10 10:03 ` juzhe.zhong
2024-01-10 10:57 ` Re：Re：[PATCH " joshua
2024-01-10 9:31 ` [PATCH " Jun Sha (Joshua)
2024-01-11 8:46 ` Jun Sha (Joshua)
2024-01-11 9:07 ` juzhe.zhong
2024-01-11 9:11 ` Re：[PATCH " joshua
2024-01-11 9:14 ` joshua
2024-01-11 9:17 ` juzhe.zhong
2024-01-11 9:21 ` Re：Re：[PATCH " joshua
2024-01-11 9:24 ` Re：[PATCH " juzhe.zhong
2024-01-11 9:29 ` Re：Re：[PATCH " joshua
2024-01-11 9:32 ` Re：[PATCH " juzhe.zhong
2024-01-11 9:38 ` Re：Re：[PATCH " joshua
2024-01-11 12:05 ` joshua
2024-01-11 12:13 ` Re：[PATCH " juzhe.zhong
2024-01-11 12:18 ` Re：Re：[PATCH " joshua
2024-01-11 12:28 ` Re：[PATCH " juzhe.zhong
2024-01-11 12:31 ` Re：Re：[PATCH " joshua
2024-01-11 12:33 ` Re：[PATCH " juzhe.zhong
2024-01-11 12:36 ` Re：Re：[PATCH " joshua
2024-01-11 9:26 ` joshua
2024-01-11 9:28 ` Re：[PATCH " juzhe.zhong
2024-01-11 9:35 ` Re：Re：[PATCH " joshua
2024-01-11 9:54 ` Re：[PATCH " joshua
2024-01-11 9:52 ` [PATCH " Jun Sha (Joshua)
2024-01-11 9:57 ` juzhe.zhong
2024-01-11 10:54 ` Re：[PATCH " joshua
2024-01-11 10:55 ` juzhe.zhong
2024-01-11 14:11 ` Re：Re：[PATCH " joshua
2024-01-11 22:59 ` Re：[PATCH " 钟居哲
2024-01-11 23:22 ` 钟居哲
2024-01-12 0:49 ` 回复：Re：[PATCH " joshua
2024-01-12 1:08 ` 回复：[PATCH " juzhe.zhong
2024-01-12 1:14 ` juzhe.zhong
2024-01-12 3:26 ` Re：Re：[PATCH " joshua
2024-01-03 2:37 ` [PATCH v4] RISC-V: Fix register overlap issue for some xtheadvector instructions Jun Sha (Joshua)
2024-01-03 7:54 ` Jun Sha (Joshua)
2024-01-10 6:02 ` [PATCH v5] " Jun Sha (Joshua)
2024-01-10 6:37 ` juzhe.zhong
2024-01-10 6:51 ` Jun Sha (Joshua)
2024-01-10 6:53 ` juzhe.zhong
2024-01-10 13:36 ` Robin Dapp
2024-01-10 13:43 ` 钟居哲
2024-01-11 2:40 ` Re：Re: " joshua
2024-01-11 2:39 ` Jun Sha (Joshua)
2024-01-11 2:46 ` juzhe.zhong
2024-01-11 8:12 ` Robin Dapp
2024-01-03 2:39 ` [PATCH v4] RISC-V: Rewrite some instructions using ASM targethook Jun Sha (Joshua)
2024-01-03 3:05 ` Kito Cheng
2024-01-03 7:55 ` Jun Sha (Joshua)
2024-01-12 3:18 ` [PATCH v5] RISC-V: Support XTheadVector extension Jun Sha (Joshua)
2024-01-12 3:20 ` [PATCH v4] RISC-V: Introduce XTheadVector as a subset of V1.0.0 Jun Sha (Joshua)
2024-01-12 7:31 ` juzhe.zhong
2024-01-18 9:50 ` Kito Cheng
2024-01-12 3:21 ` [PATCH v5] RISC-V: Adds the prefix "th." for the instructions of XTheadVector Jun Sha (Joshua)
2024-01-12 7:31 ` juzhe.zhong
2024-01-12 3:22 ` [PATCH v6] RISC-V: Handle differences between XTheadvector and Vector Jun Sha (Joshua)
2024-01-12 7:31 ` juzhe.zhong
2024-01-12 3:22 ` [PATCH v6] RISC-V: Add support for xtheadvector-specific intrinsics Jun Sha (Joshua)
2024-01-12 7:32 ` juzhe.zhong
2024-01-12 3:23 ` [PATCH v6] RISC-V: Fix register overlap issue for some xtheadvector instructions Jun Sha (Joshua)
2024-01-12 7:32 ` juzhe.zhong
2024-01-12 3:24 ` [PATCH v5] RISC-V: Rewrite some instructions using ASM targethook Jun Sha (Joshua)
2024-01-12 7:32 ` juzhe.zhong
2024-01-18 14:43 ` [PATCH v5] RISC-V: Support XTheadVector extension Christoph Müllner
2024-01-19 20:03 ` Jeff Law
[not found] <84F7577214BA46AD+202401101900269053910@rivai.ai>
2024-01-10 11:06 ` Re：Re：[PATCH v5] RISC-V: Add support for xtheadvector-specific intrinsics joshua
2024-01-10 11:08 ` joshua
2024-01-10 11:09 Re：[PATCH " juzhe.zhong
2024-01-10 11:14 ` Re：Re：[PATCH " joshua
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).