procyberian/gcc
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

re PR target/85832 ([AVX512] possible shorter code when comparing with vector of zeros)

Browse source 
PR target/85832
	* config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
	Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
	in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.

	* gcc.target/i386/avx512f-pr85832.c: New test.
	* gcc.target/i386/avx512vl-pr85832.c: New test.
	* gcc.target/i386/avx512bw-pr85832.c: New test.
	* gcc.target/i386/avx512vlbw-pr85832.c: New test.

From-SVN: r260756
This commit is contained in:
Jakub Jelinek 2018-05-25 14:36:03 +02:00 • committed by Jakub Jelinek
parent 8e846c6670
commit b55f342bdd
7 changed files with 128 additions and 9 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

7
gcc/ChangeLog
View file

@ -1,3 +1,10 @@
2018-05-25 Jakub Jelinek <jakub@redhat.com>
PR target/85832
* config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
2018-05-25 Richard Biener <rguenther@suse.de>
* tree-vect-data-refs.c (vect_find_stmt_data_reference): New

22
gcc/config/i386/sse.md
View file

@ -11207,26 +11207,30 @@
"ix86_fixup_binary_operands_no_copy (EQ, <MODE>mode, operands);")
(define_insn "<avx512>_eq<mode>3<mask_scalar_merge_name>_1"
[(set (match_operand:<avx512fmaskmode> 0 "register_operand" "=Yk")
[(set (match_operand:<avx512fmaskmode> 0 "register_operand" "=Yk,Yk")
(unspec:<avx512fmaskmode>
[(match_operand:VI12_AVX512VL 1 "nonimmediate_operand" "%v")
(match_operand:VI12_AVX512VL 2 "nonimmediate_operand" "vm")]
[(match_operand:VI12_AVX512VL 1 "vector_move_operand" "%v,v")
(match_operand:VI12_AVX512VL 2 "vector_move_operand" "vm,C")]
UNSPEC_MASKED_EQ))]
"TARGET_AVX512F && !(MEM_P (operands[1]) && MEM_P (operands[2]))"
"vpcmpeq<ssemodesuffix>\t{%2, %1, %0<mask_scalar_merge_operand3>|%0<mask_scalar_merge_operand3>, %1, %2}"
"TARGET_AVX512BW && !(MEM_P (operands[1]) && MEM_P (operands[2]))"
"@
vpcmpeq<ssemodesuffix>\t{%2, %1, %0<mask_scalar_merge_operand3>|%0<mask_scalar_merge_operand3>, %1, %2}
vptestm<ssemodesuffix>\t{%1, %1, %0<mask_scalar_merge_operand3>|%0<mask_scalar_merge_operand3>, %1, %1}"
[(set_attr "type" "ssecmp")
(set_attr "prefix_extra" "1")
(set_attr "prefix" "evex")
(set_attr "mode" "<sseinsnmode>")])
(define_insn "<avx512>_eq<mode>3<mask_scalar_merge_name>_1"
[(set (match_operand:<avx512fmaskmode> 0 "register_operand" "=Yk")
[(set (match_operand:<avx512fmaskmode> 0 "register_operand" "=Yk,Yk")
(unspec:<avx512fmaskmode>
[(match_operand:VI48_AVX512VL 1 "nonimmediate_operand" "%v")
(match_operand:VI48_AVX512VL 2 "nonimmediate_operand" "vm")]
[(match_operand:VI48_AVX512VL 1 "vector_move_operand" "%v,v")
(match_operand:VI48_AVX512VL 2 "vector_move_operand" "vm,C")]
UNSPEC_MASKED_EQ))]
"TARGET_AVX512F && !(MEM_P (operands[1]) && MEM_P (operands[2]))"
"vpcmpeq<ssemodesuffix>\t{%2, %1, %0<mask_scalar_merge_operand3>|%0<mask_scalar_merge_operand3>, %1, %2}"
"@
vpcmpeq<ssemodesuffix>\t{%2, %1, %0<mask_scalar_merge_operand3>|%0<mask_scalar_merge_operand3>, %1, %2}
vptestm<ssemodesuffix>\t{%1, %1, %0<mask_scalar_merge_operand3>|%0<mask_scalar_merge_operand3>, %1, %1}"
[(set_attr "type" "ssecmp")
(set_attr "prefix_extra" "1")
(set_attr "prefix" "evex")

8
gcc/testsuite/ChangeLog
View file

@ -1,3 +1,11 @@
2018-05-25 Jakub Jelinek <jakub@redhat.com>
PR target/85832
* gcc.target/i386/avx512f-pr85832.c: New test.
* gcc.target/i386/avx512vl-pr85832.c: New test.
* gcc.target/i386/avx512bw-pr85832.c: New test.
* gcc.target/i386/avx512vlbw-pr85832.c: New test.
2018-05-25 Bin Cheng <bin.cheng@arm.com>
PR tree-optimization/85720

19
gcc/testsuite/gcc.target/i386/avx512bw-pr85832.c Normal file
View file

@ -0,0 +1,19 @@
/* PR target/85832 */
/* { dg-do compile } */
/* { dg-options "-O2 -mavx512bw -mno-avx512vl -masm=att" } */
/* { dg-final { scan-assembler-times {\mvptestmb\M} 1 } } */
/* { dg-final { scan-assembler-times {\mvptestmw\M} 1 } } */
#include <x86intrin.h>
int
f1 (__m512i x)
{
return _mm512_cmpeq_epi8_mask (x, _mm512_setzero_si512 ());
}
int
f2 (__m512i x)
{
return _mm512_cmpeq_epi16_mask (x, _mm512_setzero_si512 ());
}

19
gcc/testsuite/gcc.target/i386/avx512f-pr85832.c Normal file
View file

@ -0,0 +1,19 @@
/* PR target/85832 */
/* { dg-do compile } */
/* { dg-options "-O2 -mavx512f -mno-avx512vl -mno-avx512bw -masm=att" } */
/* { dg-final { scan-assembler-times {\mvptestmd\M} 1 } } */
/* { dg-final { scan-assembler-times {\mvptestmq\M} 1 } } */
#include <x86intrin.h>
int
f1 (__m512i x)
{
return _mm512_cmpeq_epi32_mask (x, _mm512_setzero_si512 ());
}
int
f2 (__m512i x)
{
return _mm512_cmpeq_epi64_mask (x, _mm512_setzero_si512 ());
}

31
gcc/testsuite/gcc.target/i386/avx512vl-pr85832.c Normal file
View file

@ -0,0 +1,31 @@
/* PR target/85832 */
/* { dg-do compile } */
/* { dg-options "-O2 -mavx512vl -mno-avx512bw -masm=att" } */
/* { dg-final { scan-assembler-times {\mvptestmd\M} 2 } } */
/* { dg-final { scan-assembler-times {\mvptestmq\M} 2 } } */
#include <x86intrin.h>
int
f1 (__m256i x)
{
return _mm256_cmpeq_epi32_mask (x, _mm256_setzero_si256 ());
}
int
f2 (__m256i x)
{
return _mm256_cmpeq_epi64_mask (x, _mm256_setzero_si256 ());
}
int
f3 (__m128i x)
{
return _mm_cmpeq_epi32_mask (x, _mm_setzero_si128 ());
}
int
f4 (__m128i x)
{
return _mm_cmpeq_epi64_mask (x, _mm_setzero_si128 ());
}

31
gcc/testsuite/gcc.target/i386/avx512vlbw-pr85832.c Normal file
View file

@ -0,0 +1,31 @@
/* PR target/85832 */
/* { dg-do compile } */
/* { dg-options "-O2 -mavx512vl -mavx512bw -masm=att" } */
/* { dg-final { scan-assembler-times {\mvptestmb\M} 2 } } */
/* { dg-final { scan-assembler-times {\mvptestmw\M} 2 } } */
#include <x86intrin.h>
int
f1 (__m256i x)
{
return _mm256_cmpeq_epi8_mask (x, _mm256_setzero_si256 ());
}
int
f2 (__m256i x)
{
return _mm256_cmpeq_epi16_mask (x, _mm256_setzero_si256 ());
}
int
f3 (__m128i x)
{
return _mm_cmpeq_epi8_mask (x, _mm_setzero_si128 ());
}
int
f4 (__m128i x)
{
return _mm_cmpeq_epi16_mask (x, _mm_setzero_si128 ());
}