From: Philipp Tomsich <philipp.tomsich@vrull.eu>
To: gcc-patches@gcc.gnu.org
Cc: Vineet Gupta <vineetg@rivosinc.com>,
Palmer Dabbelt <palmer@rivosinc.com>,
Christoph Muellner <christoph.muellner@vrull.eu>,
Kito Cheng <kito.cheng@gmail.com>,
Jeff Law <jlaw@ventanamicro.com>,
Philipp Tomsich <philipp.tomsich@vrull.eu>,
Henry Brausen <henry.brausen@vrull.eu>
Subject: [PATCH 6/7] RISC-V: Support immediates in XVentanaCondOps
Date: Sat, 12 Nov 2022 22:29:42 +0100 [thread overview]
Message-ID: <20221112212943.3068249-7-philipp.tomsich@vrull.eu> (raw)
In-Reply-To: <20221112212943.3068249-1-philipp.tomsich@vrull.eu>
When if-conversion encounters sequences using immediates, the
sequences can't trivially map back onto vt.maskc/vt.maskcn (even if
benefitial) due to vt.maskc and vt.maskcn not having immediate forms.
This adds a splitter to rewrite opportunities for XVentanaCondOps that
operate on an immediate by first putting the immediate into a register
to enable the non-immediate vt.maskc/vt.maskcn instructions to operate
on the value.
Consider code, such as
long func2 (long a, long c)
{
if (c)
a = 2;
else
a = 5;
return a;
}
which will be converted to
func2:
seqz a0,a2
neg a0,a0
andi a0,a0,3
addi a0,a0,2
ret
Following this change, we generate
li a0,3
vt.maskcn a0,a0,a2
addi a0,a0,2
ret
This commit also introduces a simple unit test for if-conversion with
immediate (literal) values as the sources for simple sets in the THEN
and ELSE blocks. The test checks that Ventana's conditional mask
instruction (vt.maskc<n>) is emitted as part of the resultant branchless
instruction sequence.
gcc/ChangeLog:
* config/riscv/xventanacondops.md: Support immediates for
vt.maskc/vt.maskcn through a splitter.
gcc/testsuite/ChangeLog:
* gcc.target/riscv/xventanacondops-ifconv-imm.c: New test.
Signed-off-by: Philipp Tomsich <philipp.tomsich@vrull.eu>
Reviewed-by: Henry Brausen <henry.brausen@vrull.eu>
---
Ref #204
gcc/config/riscv/xventanacondops.md | 24 +++++++++++++++++--
.../riscv/xventanacondops-ifconv-imm.c | 19 +++++++++++++++
2 files changed, 41 insertions(+), 2 deletions(-)
create mode 100644 gcc/testsuite/gcc.target/riscv/xventanacondops-ifconv-imm.c
diff --git a/gcc/config/riscv/xventanacondops.md b/gcc/config/riscv/xventanacondops.md
index 22b4b7d103a..0e09ee91a69 100644
--- a/gcc/config/riscv/xventanacondops.md
+++ b/gcc/config/riscv/xventanacondops.md
@@ -29,6 +29,26 @@
"TARGET_XVENTANACONDOPS"
"vt.maskc<n>\t%0,%2,%1")
+;; XVentanaCondOps does not have immediate forms, so we need to do extra
+;; work to support these: if we encounter a vt.maskc/n with an immediate,
+;; we split this into a load-immediate followed by a vt.maskc/n.
+(define_split
+ [(set (match_operand:DI 0 "register_operand")
+ (and:DI (neg:DI (match_operator:DI 1 "equality_operator"
+ [(match_operand:DI 2 "register_operand")
+ (const_int 0)]))
+ (match_operand:DI 3 "immediate_operand")))
+ (clobber (match_operand:DI 4 "register_operand"))]
+ "TARGET_XVENTANACONDOPS"
+ [(set (match_dup 4) (match_dup 3))
+ (set (match_dup 0) (and:DI (neg:DI (match_dup 1))
+ (match_dup 4)))]
+{
+ /* Eliminate the clobber/temporary, if it is not needed. */
+ if (!rtx_equal_p (operands[0], operands[2]))
+ operands[4] = operands[0];
+})
+
;; Make order operators digestible to the vt.maskc<n> logic by
;; wrapping their result in a comparison against (const_int 0).
@@ -37,7 +57,7 @@
[(set (match_operand:X 0 "register_operand")
(and:X (neg:X (match_operator:X 1 "anyge_operator"
[(match_operand:X 2 "register_operand")
- (match_operand:X 3 "register_operand")]))
+ (match_operand:X 3 "arith_operand")]))
(match_operand:X 4 "register_operand")))
(clobber (match_operand:X 5 "register_operand"))]
"TARGET_XVENTANACONDOPS"
@@ -54,7 +74,7 @@
[(set (match_operand:X 0 "register_operand")
(and:X (neg:X (match_operator:X 1 "anygt_operator"
[(match_operand:X 2 "register_operand")
- (match_operand:X 3 "register_operand")]))
+ (match_operand:X 3 "arith_operand")]))
(match_operand:X 4 "register_operand")))
(clobber (match_operand:X 5 "register_operand"))]
"TARGET_XVENTANACONDOPS"
diff --git a/gcc/testsuite/gcc.target/riscv/xventanacondops-ifconv-imm.c b/gcc/testsuite/gcc.target/riscv/xventanacondops-ifconv-imm.c
new file mode 100644
index 00000000000..0012e7b669c
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/xventanacondops-ifconv-imm.c
@@ -0,0 +1,19 @@
+/* { dg-do compile } */
+/* { dg-options "-march=rv64gc_xventanacondops -mabi=lp64" } */
+/* { dg-skip-if "" { *-*-* } { "-O0" "-Og" "-Os" "-Oz" } } */
+
+/* Each function below should emit a vt.maskcn instruction */
+
+long
+foo0 (long a, long b, long c)
+{
+ if (c)
+ a = 0;
+ else
+ a = 5;
+ return a;
+}
+
+/* { dg-final { scan-assembler-times "vt.maskcn\t" 1 } } */
+/* { dg-final { scan-assembler-not "beqz\t" } } */
+/* { dg-final { scan-assembler-not "bnez\t" } } */
--
2.34.1
next prev parent reply other threads:[~2022-11-12 21:29 UTC|newest]
Thread overview: 21+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-11-12 21:29 [PATCH 0/7] RISC-V: Backend support for XVentanaCondOps/ZiCondops Philipp Tomsich
2022-11-12 21:29 ` [PATCH 1/7] RISC-V: Recognize xventanacondops extension Philipp Tomsich
2022-11-17 22:46 ` Jeff Law
2022-11-12 21:29 ` [PATCH 2/7] RISC-V: Generate vt.maskc<n> on noce_try_store_flag_mask if-conversion Philipp Tomsich
2022-11-17 22:49 ` Jeff Law
2022-11-12 21:29 ` [PATCH 3/7] RISC-V: Support noce_try_store_flag_mask as vt.maskc<n> Philipp Tomsich
2022-11-17 23:12 ` Jeff Law
2022-11-12 21:29 ` [PATCH 4/7] RISC-V: Recognize sign-extract + and cases for XVentanaCondOps Philipp Tomsich
2022-11-17 23:41 ` Jeff Law
2022-11-17 23:56 ` Palmer Dabbelt
2022-11-18 0:10 ` Philipp Tomsich
2022-11-18 14:34 ` Jeff Law
2022-11-18 14:41 ` Philipp Tomsich
2022-11-18 0:08 ` Philipp Tomsich
2022-11-12 21:29 ` [PATCH 5/7] RISC-V: Recognize bexti in negated if-conversion Philipp Tomsich
2022-11-17 23:17 ` Jeff Law
2022-11-12 21:29 ` Philipp Tomsich [this message]
2022-11-17 23:36 ` [PATCH 6/7] RISC-V: Support immediates in XVentanaCondOps Jeff Law
2022-11-12 21:29 ` [PATCH 7/7] ifcvt: add if-conversion to conditional-zero instructions Philipp Tomsich
2022-11-12 21:47 ` Andrew Pinski
2022-11-12 22:01 ` Philipp Tomsich
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20221112212943.3068249-7-philipp.tomsich@vrull.eu \
--to=philipp.tomsich@vrull.eu \
--cc=christoph.muellner@vrull.eu \
--cc=gcc-patches@gcc.gnu.org \
--cc=henry.brausen@vrull.eu \
--cc=jlaw@ventanamicro.com \
--cc=kito.cheng@gmail.com \
--cc=palmer@rivosinc.com \
--cc=vineetg@rivosinc.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
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).