This is an archive of the discontinued LLVM Phabricator instance.

Differential D20061

[Clang][AVX512] completing missing intrinsics [vextractf].
ClosedPublic

Authored by m_zuckerman on May 9 2016, 3:45 AM.

Details

Reviewers
AsafBadouh
delena
igorb
Commits
rG2564d2f5fe3b: [Clang][AVX512] completing missing intrinsics [vextractf].
rC269037: [Clang][AVX512] completing missing intrinsics [vextractf].
rL269037: [Clang][AVX512] completing missing intrinsics [vextractf].

Diff Detail

Event Timeline

m_zuckerman updated this revision to Diff 56546.May 9 2016, 3:45 AM
m_zuckerman retitled this revision from to [Clang][AVX512] completing missing intrinsics [vextractf]..
m_zuckerman updated this object.
m_zuckerman added reviewers: AsafBadouh, igorb, delena.
AsafBadouh accepted this revision.May 10 2016, 12:56 AM
AsafBadouh edited edge metadata.

LGTM

This revision is now accepted and ready to land.May 10 2016, 12:56 AM
Closed by commit rL269037: [Clang][AVX512] completing missing intrinsics [vextractf]. (authored by mzuckerm). · Explain WhyMay 10 2016, 3:21 AM
This revision was automatically updated to reflect the committed changes.

Revision Contents

PathSize
lib/
Headers/
avx512fintrin.h
avx512fintrin.h (revision 268189)
35 lines
test/
CodeGen/
avx512f-builtins.c
avx512f-builtins.c (revision 268189)
24 lines

Diff 56546

lib/Headers/avx512fintrin.h

  • This file is larger than 256 KB, so syntax highlighting is disabled by default.
Show First 20 Lines • Show All 2,581 Lines • ▼ Show 20 Lines#define _mm512_maskz_alignr_epi32( __U, __A, __B, __imm) __extension__({\
(__m512i) __builtin_ia32_alignd512_mask ((__v16si) __A,\ (__m512i) __builtin_ia32_alignd512_mask ((__v16si) __A,\
(__v16si) __B, __imm,\ (__v16si) __B, __imm,\
(__v16si) _mm512_setzero_si512 (),\ (__v16si) _mm512_setzero_si512 (),\
(__mmask16) __U);\ (__mmask16) __U);\
}) })
/* Vector Extract */ /* Vector Extract */
#define _mm512_extractf64x4_pd(A, I) __extension__ ({ \ #define _mm512_extractf64x4_pd(A, I) __extension__ ({ \
(__m256d) \ (__m256d) __builtin_ia32_extractf64x4_mask((__v8df)(__m512d)(A), \
__builtin_ia32_extractf64x4_mask((__v8df)(__m512d)(A), \
(I), \ (I), \
(__v4df)_mm256_setzero_si256(), \ (__v4df)_mm256_setzero_si256(), \
(__mmask8) -1); }) (__mmask8) -1); })
#define _mm512_mask_extractf64x4_pd( __W, __U, __A, __imm) __extension__ ({\
(__m256d) __builtin_ia32_extractf64x4_mask ((__v8df) __A, \
__imm,\
(__v4df) __W,\
(__mmask8) __U);\
})
#define _mm512_maskz_extractf64x4_pd( __U, __A, __imm) __extension__ ({\
(__m256d) __builtin_ia32_extractf64x4_mask ((__v8df) __A,\
__imm,\
(__v4df)\
_mm256_setzero_pd (),\
(__mmask8) __U);\
})
#define _mm512_extractf32x4_ps(A, I) __extension__ ({ \ #define _mm512_extractf32x4_ps(A, I) __extension__ ({ \
(__m128) \ (__m128) __builtin_ia32_extractf32x4_mask((__v16sf)(__m512)(A), \
__builtin_ia32_extractf32x4_mask((__v16sf)(__m512)(A), \
(I), \ (I), \
(__v4sf)_mm_setzero_ps(), \ (__v4sf)_mm_setzero_ps(), \
(__mmask8) -1); }) (__mmask8) -1); })
#define _mm512_mask_extractf32x4_ps( __W, __U, __A, __imm) __extension__ ({\
(__m128) __builtin_ia32_extractf32x4_mask ((__v16sf) __A,\
__imm,\
(__v4sf) __W,\
(__mmask8) __U);\
})
#define _mm512_maskz_extractf32x4_ps( __U, __A, __imm) __extension__ ({\
(__m128) __builtin_ia32_extractf32x4_mask ((__v16sf) __A,\
__imm,\
(__v4sf)\
_mm_setzero_ps (),\
(__mmask8) __U);\
})
/* Vector Blend */ /* Vector Blend */
static __inline __m512d __DEFAULT_FN_ATTRS static __inline __m512d __DEFAULT_FN_ATTRS
_mm512_mask_blend_pd(__mmask8 __U, __m512d __A, __m512d __W) _mm512_mask_blend_pd(__mmask8 __U, __m512d __A, __m512d __W)
{ {
return (__m512d) __builtin_ia32_blendmpd_512_mask ((__v8df) __A, return (__m512d) __builtin_ia32_blendmpd_512_mask ((__v8df) __A,
(__v8df) __W, (__v8df) __W,
(__mmask8) __U); (__mmask8) __U);
▲ Show 20 Lines • Show All 5,076 Lines • Show Last 20 Lines

test/CodeGen/avx512f-builtins.c

Show First 20 Lines • Show All 836 Lines • ▼ Show 20 Lines
__m256d test_mm512_extractf64x4_pd(__m512d a) __m256d test_mm512_extractf64x4_pd(__m512d a)
{ {
// CHECK-LABEL: @test_mm512_extractf64x4_pd // CHECK-LABEL: @test_mm512_extractf64x4_pd
// CHECK: @llvm.x86.avx512.mask.vextractf64x4.512 // CHECK: @llvm.x86.avx512.mask.vextractf64x4.512
return _mm512_extractf64x4_pd(a, 1); return _mm512_extractf64x4_pd(a, 1);
} }
__m256d test_mm512_mask_extractf64x4_pd(__m256d __W,__mmask8 __U,__m512d __A){
//CHECK-LABLE:@test_mm512_mask_extractf64x4_pd
//CHECL:@llvm.x86.avx512.mask.vextractf64x4.512
return _mm512_mask_extractf64x4_pd( __W, __U, __A, 1);
}
__m256d test_mm512_maskz_extractf64x4_pd(__mmask8 __U,__m512d __A){
//CHECK-LABLE:@test_mm512_maskz_extractf64x4_pd
//CHECL:@llvm.x86.avx512.mask.vextractf64x4.512
return _mm512_maskz_extractf64x4_pd( __U, __A, 1);
}
__m128 test_mm512_extractf32x4_ps(__m512 a) __m128 test_mm512_extractf32x4_ps(__m512 a)
{ {
// CHECK-LABEL: @test_mm512_extractf32x4_ps // CHECK-LABEL: @test_mm512_extractf32x4_ps
// CHECK: @llvm.x86.avx512.mask.vextractf32x4.512 // CHECK: @llvm.x86.avx512.mask.vextractf32x4.512
return _mm512_extractf32x4_ps(a, 1); return _mm512_extractf32x4_ps(a, 1);
} }
__m128 test_mm512_mask_extractf32x4_ps(__m128 __W, __mmask8 __U,__m512d __A){
//CHECK-LABLE:@test_mm512_mask_extractf32x4_ps
//CHECL: @llvm.x86.avx512.mask.vextractf32x4.512
return _mm512_mask_extractf32x4_ps( __W, __U, __A, 1);
}
__m128 test_mm512_maskz_extractf32x4_ps( __mmask8 __U,__m512d __A){
//CHECK-LABLE:@test_mm512_maskz_extractf32x4_ps
//CHECL: @llvm.x86.avx512.mask.vextractf32x4.512
return _mm512_maskz_extractf32x4_ps( __U, __A, 1);
}
__mmask16 test_mm512_cmpeq_epu32_mask(__m512i __a, __m512i __b) { __mmask16 test_mm512_cmpeq_epu32_mask(__m512i __a, __m512i __b) {
// CHECK-LABEL: @test_mm512_cmpeq_epu32_mask // CHECK-LABEL: @test_mm512_cmpeq_epu32_mask
// CHECK: @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32> {{.*}}, <16 x i32> {{.*}}, i32 0, i16 -1) // CHECK: @llvm.x86.avx512.mask.ucmp.d.512(<16 x i32> {{.*}}, <16 x i32> {{.*}}, i32 0, i16 -1)
return (__mmask16)_mm512_cmpeq_epu32_mask(__a, __b); return (__mmask16)_mm512_cmpeq_epu32_mask(__a, __b);
} }
__mmask16 test_mm512_mask_cmpeq_epu32_mask(__mmask16 __u, __m512i __a, __m512i __b) { __mmask16 test_mm512_mask_cmpeq_epu32_mask(__mmask16 __u, __m512i __a, __m512i __b) {
// CHECK-LABEL: @test_mm512_mask_cmpeq_epu32_mask // CHECK-LABEL: @test_mm512_mask_cmpeq_epu32_mask
▲ Show 20 Lines • Show All 4,476 Lines • Show Last 20 Lines