Re: [PATCH v1] RISC-V: Support RVV VFWMUL rounding mode intrinsic API

[钟居哲 <juzhe.zhong@rivai.ai>]

[-- Attachment #1: Type: text/plain, Size: 6273 bytes --]

LGTM



juzhe.zhong@rivai.ai
 
From: pan2.li
Date: 2023-08-03 13:28
To: gcc-patches
CC: juzhe.zhong; pan2.li; yanzhang.wang; kito.cheng
Subject: [PATCH v1] RISC-V: Support RVV VFWMUL rounding mode intrinsic API
From: Pan Li <pan2.li@intel.com>
 
This patch would like to support the rounding mode API for the
VFWMUL for the below samples.
 
* __riscv_vfwmul_vv_f64m2_rm
* __riscv_vfwmul_vv_f64m2_rm_m
* __riscv_vfwmul_vf_f64m2_rm
* __riscv_vfwmul_vf_f64m2_rm_m
 
Signed-off-by: Pan Li <pan2.li@intel.com>
 
gcc/ChangeLog:
 
* config/riscv/riscv-vector-builtins-bases.cc
(vfwmul_frm_obj): New declaration.
(vfwmul_frm): Ditto.
* config/riscv/riscv-vector-builtins-bases.h:
(vfwmul_frm): Ditto.
* config/riscv/riscv-vector-builtins-functions.def
(vfwmul_frm): New function definition.
* config/riscv/vector.md: (frm_mode) Add vfwmul to frm_mode.
 
gcc/testsuite/ChangeLog:
 
* gcc.target/riscv/rvv/base/float-point-widening-mul.c: New test.
---
.../riscv/riscv-vector-builtins-bases.cc      |  3 ++
.../riscv/riscv-vector-builtins-bases.h       |  1 +
.../riscv/riscv-vector-builtins-functions.def |  2 +
gcc/config/riscv/vector.md                    |  2 +-
.../riscv/rvv/base/float-point-widening-mul.c | 44 +++++++++++++++++++
5 files changed, 51 insertions(+), 1 deletion(-)
create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/base/float-point-widening-mul.c
 
diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.cc b/gcc/config/riscv/riscv-vector-builtins-bases.cc
index 95ec9ccb481..8d689f0c935 100644
--- a/gcc/config/riscv/riscv-vector-builtins-bases.cc
+++ b/gcc/config/riscv/riscv-vector-builtins-bases.cc
@@ -321,6 +321,7 @@ public:
/* Implements below instructions for frm
    - vfwadd
    - vfwsub
+   - vfwmul
*/
template<rtx_code CODE>
class widen_binop_frm : public function_base
@@ -2112,6 +2113,7 @@ static CONSTEXPR const binop_frm<DIV> vfdiv_frm_obj;
static CONSTEXPR const reverse_binop<DIV> vfrdiv_obj;
static CONSTEXPR const reverse_binop_frm<DIV> vfrdiv_frm_obj;
static CONSTEXPR const widen_binop<MULT> vfwmul_obj;
+static CONSTEXPR const widen_binop_frm<MULT> vfwmul_frm_obj;
static CONSTEXPR const vfmacc vfmacc_obj;
static CONSTEXPR const vfnmsac vfnmsac_obj;
static CONSTEXPR const vfmadd vfmadd_obj;
@@ -2346,6 +2348,7 @@ BASE (vfdiv_frm)
BASE (vfrdiv)
BASE (vfrdiv_frm)
BASE (vfwmul)
+BASE (vfwmul_frm)
BASE (vfmacc)
BASE (vfnmsac)
BASE (vfmadd)
diff --git a/gcc/config/riscv/riscv-vector-builtins-bases.h b/gcc/config/riscv/riscv-vector-builtins-bases.h
index f35fd3d27cf..2d2b52a312c 100644
--- a/gcc/config/riscv/riscv-vector-builtins-bases.h
+++ b/gcc/config/riscv/riscv-vector-builtins-bases.h
@@ -158,6 +158,7 @@ extern const function_base *const vfdiv_frm;
extern const function_base *const vfrdiv;
extern const function_base *const vfrdiv_frm;
extern const function_base *const vfwmul;
+extern const function_base *const vfwmul_frm;
extern const function_base *const vfmacc;
extern const function_base *const vfnmsac;
extern const function_base *const vfmadd;
diff --git a/gcc/config/riscv/riscv-vector-builtins-functions.def b/gcc/config/riscv/riscv-vector-builtins-functions.def
index e7e6c7d8ed8..d43b33ded17 100644
--- a/gcc/config/riscv/riscv-vector-builtins-functions.def
+++ b/gcc/config/riscv/riscv-vector-builtins-functions.def
@@ -328,6 +328,8 @@ DEF_RVV_FUNCTION (vfrdiv_frm, alu_frm, full_preds, f_vvf_ops)
// 13.5. Vector Widening Floating-Point Multiply
DEF_RVV_FUNCTION (vfwmul, alu, full_preds, f_wvv_ops)
DEF_RVV_FUNCTION (vfwmul, alu, full_preds, f_wvf_ops)
+DEF_RVV_FUNCTION (vfwmul_frm, alu_frm, full_preds, f_wvv_ops)
+DEF_RVV_FUNCTION (vfwmul_frm, alu_frm, full_preds, f_wvf_ops)
// 13.6. Vector Single-Width Floating-Point Fused Multiply-Add Instructions
DEF_RVV_FUNCTION (vfmacc, alu, full_preds, f_vvvv_ops)
diff --git a/gcc/config/riscv/vector.md b/gcc/config/riscv/vector.md
index 4b6c3859947..750b2de8df9 100644
--- a/gcc/config/riscv/vector.md
+++ b/gcc/config/riscv/vector.md
@@ -866,7 +866,7 @@ (define_attr "vxrm_mode" "rnu,rne,rdn,rod,none"
;; Defines rounding mode of an floating-point operation.
(define_attr "frm_mode" "rne,rtz,rdn,rup,rmm,dyn,dyn_exit,dyn_call,none"
-  (cond [(eq_attr "type" "vfalu,vfwalu,vfmul,vfdiv")
+  (cond [(eq_attr "type" "vfalu,vfwalu,vfmul,vfdiv,vfwmul")
          (cond
   [(match_test "INTVAL (operands[9]) == riscv_vector::FRM_RNE")
    (const_string "rne")
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/base/float-point-widening-mul.c b/gcc/testsuite/gcc.target/riscv/rvv/base/float-point-widening-mul.c
new file mode 100644
index 00000000000..893fa866a31
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/base/float-point-widening-mul.c
@@ -0,0 +1,44 @@
+/* { dg-do compile } */
+/* { dg-options "-march=rv64gcv -mabi=lp64 -O3 -Wno-psabi" } */
+
+#include "riscv_vector.h"
+
+typedef float float32_t;
+
+vfloat64m2_t
+test_vfwmul_vv_f32m1_rm (vfloat32m1_t op1, vfloat32m1_t op2, size_t vl) {
+  return __riscv_vfwmul_vv_f64m2_rm (op1, op2, 0, vl);
+}
+
+vfloat64m2_t
+test_vfwmul_vv_f32m1_rm_m (vbool32_t mask, vfloat32m1_t op1, vfloat32m1_t op2,
+    size_t vl) {
+  return __riscv_vfwmul_vv_f64m2_rm_m (mask, op1, op2, 1, vl);
+}
+
+vfloat64m2_t
+test_vfwmul_vf_f32m1_rm (vfloat32m1_t op1, float32_t op2, size_t vl) {
+  return __riscv_vfwmul_vf_f64m2_rm (op1, op2, 2, vl);
+}
+
+vfloat64m2_t
+test_vfwmul_vf_f32m1_rm_m (vbool32_t mask, vfloat32m1_t op1, float32_t op2,
+    size_t vl) {
+  return __riscv_vfwmul_vf_f64m2_rm_m (mask, op1, op2, 3, vl);
+}
+
+vfloat64m2_t
+test_vfwmul_vv_f32m1 (vfloat32m1_t op1, vfloat32m1_t op2, size_t vl) {
+  return __riscv_vfwmul_vv_f64m2 (op1, op2, vl);
+}
+
+vfloat64m2_t
+test_vfwmul_vv_f32m1_m (vbool32_t mask, vfloat32m1_t op1, vfloat32m1_t op2,
+ size_t vl) {
+  return __riscv_vfwmul_vv_f64m2_m (mask, op1, op2, vl);
+}
+
+/* { dg-final { scan-assembler-times {vfwmul\.[vw][vf]\s+v[0-9]+,\s*v[0-9]+,\s*[fav]+[0-9]+} 6 } } */
+/* { dg-final { scan-assembler-times {frrm\s+[axs][0-9]+} 4 } } */
+/* { dg-final { scan-assembler-times {fsrm\s+[axs][0-9]+} 4 } } */
+/* { dg-final { scan-assembler-times {fsrmi\s+[01234]} 4 } } */
-- 
2.34.1