OpenE2K/gcc
2
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

i386: Add AVX512 unaligned intrinsics

Browse Source 
__m512i _mm512_loadu_epi64( void * sa);
void _mm512_storeu_epi64(void * d, __m512i a);
__m512i _mm512_loadu_epi32( void * sa);
void _mm512_storeu_epi32(void * d, __m512i a);
void _mm256_storeu_epi64(void * d, __m256i a);
void _mm_storeu_epi64(void * d, __m128i a);
void _mm256_storeu_epi32(void * d, __m256i a);
void _mm_storeu_epi32(void * d, __m128i a);

Tested on x86-64.

2019-07-11  Sunil K Pandey  <sunil.k.pandey@intel.com>

gcc/

	PR target/90980
	* config/i386/avx512fintrin.h (_mm512_loadu_epi64): New.
	(_mm512_storeu_epi64): Likewise.
	(_mm512_loadu_epi32): Likewise.
	(_mm512_storeu_epi32): Likewise.
	* config/i386/avx512vlintrin.h (_mm256_storeu_epi64): New.
	(_mm_storeu_epi64): Likewise.
	(_mm256_storeu_epi32): Likewise.
	(_mm_storeu_epi32): Likewise.

gcc/testsuite/

	PR target/90980
	* gcc.target/i386/pr90980-1.c: New test.
	* gcc.target/i386/pr90980-2.c: Likewise.
	* gcc.target/i386/pr90980-3.c: Likewise.

From-SVN: r273416
This commit is contained in:
Sunil K Pandey 2019-07-11 19:09:37 +00:00 committed by H.J. Lu
parent 998175f646
commit 4c98bdadcb
7 changed files with 129 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
gcc/ChangeLog
View File

@ -1,3 +1,15 @@
2019-07-11 Sunil K Pandey <sunil.k.pandey@intel.com>
PR target/90980
* config/i386/avx512fintrin.h (_mm512_loadu_epi64): New.
(_mm512_storeu_epi64): Likewise.
(_mm512_loadu_epi32): Likewise.
(_mm512_storeu_epi32): Likewise.
* config/i386/avx512vlintrin.h (_mm256_storeu_epi64): New.
(_mm_storeu_epi64): Likewise.
(_mm256_storeu_epi32): Likewise.
(_mm_storeu_epi32): Likewise.
2019-07-11 Segher Boessenkool <segher@kernel.crashing.org>
* config/rs6000/rs6000-logue.c: Add Modula-2 to comment.

28
gcc/config/i386/avx512fintrin.h
View File

@ -6350,6 +6350,13 @@ _mm_mask_store_sd (double *__P, __mmask8 __U, __m128d __A)
__builtin_ia32_storesd_mask (__P, (__v2df) __A, (__mmask8) __U);
}
extern __inline __m512i
__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
_mm512_loadu_epi64 (void const *__P)
{
return *(__m512i_u *) __P;
}
extern __inline __m512i
__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
_mm512_mask_loadu_epi64 (__m512i __W, __mmask8 __U, void const *__P)
@ -6369,6 +6376,13 @@ _mm512_maskz_loadu_epi64 (__mmask8 __U, void const *__P)
(__mmask8) __U);
}
extern __inline void
__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
_mm512_storeu_epi64 (void *__P, __m512i __A)
{
*(__m512i_u *) __P = (__m512i_u) __A;
}
extern __inline void
__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
_mm512_mask_storeu_epi64 (void *__P, __mmask8 __U, __m512i __A)
@ -6384,6 +6398,13 @@ _mm512_loadu_si512 (void const *__P)
return *(__m512i_u *)__P;
}
extern __inline __m512i
__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
_mm512_loadu_epi32 (void const *__P)
{
return *(__m512i_u *) __P;
}
extern __inline __m512i
__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
_mm512_mask_loadu_epi32 (__m512i __W, __mmask16 __U, void const *__P)
@ -6410,6 +6431,13 @@ _mm512_storeu_si512 (void *__P, __m512i __A)
*(__m512i_u *)__P = __A;
}
extern __inline void
__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
_mm512_storeu_epi32 (void *__P, __m512i __A)
{
*(__m512i_u *) __P = (__m512i_u) __A;
}
extern __inline void
__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
_mm512_mask_storeu_epi32 (void *__P, __mmask16 __U, __m512i __A)

28
gcc/config/i386/avx512vlintrin.h
View File

@ -755,6 +755,13 @@ _mm_maskz_loadu_epi64 (__mmask8 __U, void const *__P)
(__mmask8) __U);
}
extern __inline void
__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
_mm256_storeu_epi64 (void *__P, __m256i __A)
{
*(__m256i_u *) __P = (__m256i_u) __A;
}
extern __inline void
__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
_mm256_mask_storeu_epi64 (void *__P, __mmask8 __U, __m256i __A)
@ -764,6 +771,13 @@ _mm256_mask_storeu_epi64 (void *__P, __mmask8 __U, __m256i __A)
(__mmask8) __U);
}
extern __inline void
__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
_mm_storeu_epi64 (void *__P, __m128i __A)
{
*(__m128i_u *) __P = (__m128i_u) __A;
}
extern __inline void
__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
_mm_mask_storeu_epi64 (void *__P, __mmask8 __U, __m128i __A)
@ -811,6 +825,13 @@ _mm_maskz_loadu_epi32 (__mmask8 __U, void const *__P)
(__mmask8) __U);
}
extern __inline void
__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
_mm256_storeu_epi32 (void *__P, __m256i __A)
{
*(__m256i_u *) __P = (__m256i_u) __A;
}
extern __inline void
__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
_mm256_mask_storeu_epi32 (void *__P, __mmask8 __U, __m256i __A)
@ -820,6 +841,13 @@ _mm256_mask_storeu_epi32 (void *__P, __mmask8 __U, __m256i __A)
(__mmask8) __U);
}
extern __inline void
__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
_mm_storeu_epi32 (void *__P, __m128i __A)
{
*(__m128i_u *) __P = (__m128i_u) __A;
}
extern __inline void
__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
_mm_mask_storeu_epi32 (void *__P, __mmask8 __U, __m128i __A)

7
gcc/testsuite/ChangeLog
View File

@ -1,3 +1,10 @@
2019-07-11 Sunil K Pandey <sunil.k.pandey@intel.com>
PR target/90980
* gcc.target/i386/pr90980-1.c: New test.
* gcc.target/i386/pr90980-2.c: Likewise.
* gcc.target/i386/pr90980-3.c: Likewise.
2019-07-11 Yannick Moy <moy@adacore.com>
* gnat.dg/loop_entry1.adb: New testcase.

17
gcc/testsuite/gcc.target/i386/pr90980-1.c Normal file
View File

@ -0,0 +1,17 @@
/* { dg-do compile } */
/* { dg-options "-march=skylake-avx512 -O2" } */
/* { dg-final { scan-assembler-times "(?:vmovups|vmovdqu)\[ \\t\]+\[^\{\n\]*%xmm\[0-9\]+\[^\n\]*\\)(?:\n|\[ \\t\]+#)" 2 } } */
#include <immintrin.h>
int *a;
long long *b;
volatile __m128i xx;
volatile __m128i xx1;
void extern
avx512vl_test (void)
{
_mm_storeu_epi32 (a, xx);
_mm_storeu_epi64 (b, xx1);
}

17
gcc/testsuite/gcc.target/i386/pr90980-2.c Normal file
View File

@ -0,0 +1,17 @@
/* { dg-do compile } */
/* { dg-options "-march=skylake-avx512 -O2" } */
/* { dg-final { scan-assembler-times "vmovdqu\[0-9\]*\[ \\t\]+\[^\{\n\]*%ymm\[0-9\]+\[^\n\]*\\)(?:\n|\[ \\t\]+#)" 2 } } */
#include <immintrin.h>
int *a;
long long *b;
volatile __m256i yy;
volatile __m256i yy1;
void extern
avx512vl_test (void)
{
_mm256_storeu_epi32 (a, yy);
_mm256_storeu_epi64 (b, yy1);
}

20
gcc/testsuite/gcc.target/i386/pr90980-3.c Normal file
View File

@ -0,0 +1,20 @@
/* { dg-do compile } */
/* { dg-options "-march=skylake-avx512 -O2" } */
/* { dg-final { scan-assembler-times "vmovdqu64\[ \\t\]+\[^\{\n\]*\\)\[^\n\]*%zmm\[0-9\]+(?:\n|\[ \\t\]+#)" 2 } } */
/* { dg-final { scan-assembler-times "vmovdqu64\[ \\t\]+\[^\{\n\]*%zmm\[0-9\]+\[^\n\]*\\)(?:\n|\[ \\t\]+#)" 2 } } */
#include <immintrin.h>
int *a;
long long *b;
volatile __m512i zz;
volatile __m512i zz1;
void extern
avx512f_test (void)
{
zz = _mm512_loadu_epi32 (a);
_mm512_storeu_epi32 (a, zz);
zz1 = _mm512_loadu_epi64 (b);
_mm512_storeu_epi64 (b, zz1);
}