From 7e59a44e079f6f4c9d39d625aaa3c5023cb545aa Mon Sep 17 00:00:00 2001
From: liuhongt <hongtao.liu@intel.com>
Date: Fri, 16 Apr 2021 15:32:14 +0800
Subject: [PATCH] [i386] Add pre-reload splitter for blow transformation.
- vpxor %xmm1, %xmm1, %xmm1
- vpcmpd $0, %zmm1, %zmm0, %k0
+ vptestnmd %zmm0, %zmm0, %k0
vpmovm2d zmm0, k0
gcc/ChangeLog:
PR target/98348
* config/i386/sse.md (*<avx512>_eq<mode>3): New pre_reload
splitter.
(*<avx512>_eq<mode>3): Changed from combine splitter to
pre-reload splitter.
gcc/testsuite/ChangeLog:
PR target/98348
* gcc.target/i386/pr98348.c: New test.
---
gcc/config/i386/sse.md | 23 +++++++++-
gcc/testsuite/gcc.target/i386/pr98348.c | 57 +++++++++++++++++++++++++
2 files changed, 78 insertions(+), 2 deletions(-)
create mode 100644 gcc/testsuite/gcc.target/i386/pr98348.c
@@ -3273,7 +3273,7 @@ (define_int_attr pcmp_signed_mask
;; For signed comparison, handle EQ 0: NEQ 4,
;; for unsigned comparison extra handle LE:2, NLE:6, equivalent to EQ and NEQ.
-(define_split
+(define_insn_and_split "*<avx512>_ucmp<mode>3"
[(set (match_operand:<avx512fmaskmode> 0 "register_operand")
(unspec:<avx512fmaskmode>
[(us_minus:VI12_AVX512VL
@@ -3282,9 +3282,11 @@ (define_split
(match_operand:VI12_AVX512VL 3 "const0_operand")
(match_operand:SI 4 "const_0_to_7_operand")]
UNSPEC_PCMP_ITER))]
- "TARGET_AVX512BW
+ "TARGET_AVX512BW && ix86_pre_reload_split ()
&& ix86_binary_operator_ok (US_MINUS, <MODE>mode, operands)
&& (INTVAL (operands[4]) & <pcmp_signed_mask>) == 0"
+ "#"
+ "&& 1"
[(const_int 0)]
{
/* LE: 2, NLT: 5, NLE: 6, LT: 1 */
@@ -13026,6 +13028,23 @@ (define_insn "<avx512>_eq<mode>3<mask_scalar_merge_name>_1"
(set_attr "prefix" "evex")
(set_attr "mode" "<sseinsnmode>")])
+(define_insn_and_split "*<avx512>_eq<mode>3"
+ [(set (match_operand:<avx512fmaskmode> 0 "register_operand")
+ (unspec:<avx512fmaskmode>
+ [(match_operand:VI_AVX512BW 1 "nonimm_or_0_operand")
+ (match_operand:VI_AVX512BW 2 "nonimm_or_0_operand")
+ (const_int 0)]
+ UNSPEC_PCMP_ITER))]
+ "TARGET_AVX512F && ix86_pre_reload_split ()
+ && (operands[1] == CONST0_RTX (<MODE>mode)
+ || operands[2] == CONST0_RTX (<MODE>mode))"
+ "#"
+ "&& 1"
+ [(set (match_dup 0)
+ (unspec:<avx512fmaskmode>
+ [(match_dup 1) (match_dup 2)]
+ UNSPEC_MASKED_EQ))])
+
(define_insn "*sse4_1_eqv2di3"
[(set (match_operand:V2DI 0 "register_operand" "=Yr,*x,x")
(eq:V2DI
new file mode 100644
@@ -0,0 +1,57 @@
+/* PR target/98348 */
+/* { dg-do compile } */
+/* { dg-options "-mavx512bw -O2" } */
+/* { dg-final { scan-assembler-times "vptestnm\[bwdq\]" "8" } } */
+/* { dg-final { scan-assembler-not "pcmp" } } */
+/* { dg-final { scan-assembler-not "pxor" } } */
+
+
+typedef int v16si __attribute__ ((vector_size(64)));
+typedef short v32hi __attribute__ ((vector_size(64)));
+typedef char v64qi __attribute__ ((vector_size(64)));
+typedef long long v8di __attribute__ ((vector_size(64)));
+typedef unsigned int v16usi __attribute__ ((vector_size(64)));
+typedef unsigned short v32uhi __attribute__ ((vector_size(64)));
+typedef unsigned char v64uqi __attribute__ ((vector_size(64)));
+typedef unsigned long long v8udi __attribute__ ((vector_size(64)));
+
+
+v8di
+compareq(v8di x) {
+ return x == 0;
+}
+
+v16si
+compared(v16si x) {
+ return x == 0;
+}
+
+v32hi
+comparew(v32hi x) {
+ return x == 0;
+}
+
+v64qi
+compareb(v64qi x) {
+ return x == 0;
+}
+
+v8udi
+compareuq(v8udi x) {
+ return x == 0;
+}
+
+v16usi
+compareud(v16usi x) {
+ return x == 0;
+}
+
+v32uhi
+compareuw(v32uhi x) {
+ return x == 0;
+}
+
+v64uqi
+compareub(v64uqi x) {
+ return x == 0;
+}
--
2.18.1
Hi: This patch is about to add a pre-reload splitter to transform vpcmpeqd with a zero operand to vptestnmd, which could save a vpxor instruction. .i.e - vpxor %xmm1, %xmm1, %xmm1 - vpcmpd $0, %zmm1, %zmm0, %k0 + vptestnmd %zmm0, %zmm0, %k0 vpmovm2d zmm0, k0 Bootstrapped and regtested on x86-64_iinux-gnu{-m32,}. gcc/ChangeLog: PR target/98348 * config/i386/sse.md (*<avx512>_eq<mode>3): New pre_reload splitter. (*<avx512>_eq<mode>3): Changed from combine splitter to pre-reload splitter. gcc/testsuite/ChangeLog: PR target/98348 * gcc.target/i386/pr98348.c: New test.