This is an archive of the discontinued LLVM Phabricator instance.

Differential D20810

[Clang][Intrinsics][avx512] Continue Adding round cvt to clang
ClosedPublic

Authored by m_zuckerman on May 31 2016, 5:56 AM.

Details

Reviewers
AsafBadouh
delena
igorb
Commits
rG6170c15fc68d: [Clang][Intrinsics][avx512] Continue Adding round cvt to clang
rC271398: [Clang][Intrinsics][avx512] Continue Adding round cvt to clang
rL271398: [Clang][Intrinsics][avx512] Continue Adding round cvt to clang

Diff Detail

Event Timeline

m_zuckerman updated this revision to Diff 59048.May 31 2016, 5:56 AM
m_zuckerman retitled this revision from to [Clang][Intrinsics][avx512] Continue Adding round cvt to clang.
m_zuckerman updated this object.
m_zuckerman added reviewers: AsafBadouh, igorb, delena.
m_zuckerman added a subscriber: cfe-commits.
AsafBadouh accepted this revision.Jun 1 2016, 6:41 AM
AsafBadouh edited edge metadata.

minor fixes, LGTM

lib/Headers/avx512fintrin.h
3658

please add (__mmask16)

3669

please add casting to return type (__m256i)

same for the rest

test/CodeGen/avx512f-builtins.c
3121

remove empty line

3129

remove empty line

This revision is now accepted and ready to land.Jun 1 2016, 6:41 AM
Closed by commit rL271398: [Clang][Intrinsics][avx512] Continue Adding round cvt to clang (authored by mzuckerm). · Explain WhyJun 1 2016, 7:48 AM
This revision was automatically updated to reflect the committed changes.

Revision Contents

PathSize
lib/
Headers/
64 lines
test/
CodeGen/
65 lines

Diff 59048

lib/Headers/avx512fintrin.h

  • This file is larger than 256 KB, so syntax highlighting is disabled by default.
Show First 20 Lines • Show All 3,413 Lines • ▼ Show 20 Lines
#define _mm512_cmp_pd_mask(A, B, P) \ #define _mm512_cmp_pd_mask(A, B, P) \
_mm512_cmp_round_pd_mask((A), (B), (P), _MM_FROUND_CUR_DIRECTION) _mm512_cmp_round_pd_mask((A), (B), (P), _MM_FROUND_CUR_DIRECTION)
#define _mm512_mask_cmp_pd_mask(U, A, B, P) \ #define _mm512_mask_cmp_pd_mask(U, A, B, P) \
_mm512_mask_cmp_round_pd_mask((U), (A), (B), (P), _MM_FROUND_CUR_DIRECTION) _mm512_mask_cmp_round_pd_mask((U), (A), (B), (P), _MM_FROUND_CUR_DIRECTION)
/* Conversion */ /* Conversion */
#define _mm512_cvtt_roundps_epu32( __A, __R) __extension__ ({ \
__builtin_ia32_cvttps2udq512_mask ((__v16sf)( __A),\
(__v16si)\
_mm512_undefined_epi32 (),\
(__mmask16) -1,( __R));\
})
#define _mm512_mask_cvtt_roundps_epu32( __W, __U, __A, __R) __extension__ ({ \
__builtin_ia32_cvttps2udq512_mask ((__v16sf)( __A),\
(__v16si)( __W),\
(__mmask16)( __U),( __R));\
})
#define _mm512_maskz_cvtt_roundps_epu32( __U, __A, __R) __extension__ ({ \
__builtin_ia32_cvttps2udq512_mask ((__v16sf)( __A),\
(__v16si)\
_mm512_setzero_si512 (),\
(__mmask16)( __U),( __R));\
})
static __inline __m512i __DEFAULT_FN_ATTRS static __inline __m512i __DEFAULT_FN_ATTRS
_mm512_cvttps_epu32(__m512 __A) _mm512_cvttps_epu32(__m512 __A)
{ {
return (__m512i) __builtin_ia32_cvttps2udq512_mask ((__v16sf) __A, return (__m512i) __builtin_ia32_cvttps2udq512_mask ((__v16sf) __A,
(__v16si) (__v16si)
_mm512_setzero_si512 (), _mm512_setzero_si512 (),
(__mmask16) -1, (__mmask16) -1,
_MM_FROUND_CUR_DIRECTION); _MM_FROUND_CUR_DIRECTION);
▲ Show 20 Lines • Show All 194 Lines • ▼ Show 20 Lines
_mm512_maskz_cvtpd_ps (__mmask8 __U, __m512d __A) _mm512_maskz_cvtpd_ps (__mmask8 __U, __m512d __A)
{ {
return (__m256) __builtin_ia32_cvtpd2ps512_mask ((__v8df) __A, return (__m256) __builtin_ia32_cvtpd2ps512_mask ((__v8df) __A,
(__v8sf) _mm256_setzero_ps (), (__v8sf) _mm256_setzero_ps (),
(__mmask8) __U, (__mmask8) __U,
_MM_FROUND_CUR_DIRECTION); _MM_FROUND_CUR_DIRECTION);
} }
#define _mm512_cvt_roundps_ph( __A, __I) __extension__ ({ \
__builtin_ia32_vcvtps2ph512_mask ((__v16sf)( __A),\
( __I),\
(__v16hi)\
_mm256_undefined_si256 (),\
-1);\
AsafBadouhUnsubmitted
Not Done
ReplyInline Actions

please add (__mmask16)

AsafBadouh: please add (__mmask16)
})
#define _mm512_mask_cvt_roundps_ph( __U, __W, __A, __I) __extension__ ({ \
__builtin_ia32_vcvtps2ph512_mask ((__v16sf)( __A),\
( __I),\
(__v16hi)( __U),\
(__mmask16)( __W));\
})
#define _mm512_maskz_cvt_roundps_ph( __W, __A, __I) __extension__ ({ \
__builtin_ia32_vcvtps2ph512_mask ((__v16sf)( __A),\
AsafBadouhUnsubmitted
Not Done
ReplyInline Actions

please add casting to return type (__m256i)

same for the rest

AsafBadouh: please add casting to return type (__m256i) same for the rest
( __I),\
(__v16hi)\
_mm256_setzero_si256 (),\
(__mmask16)( __W));\
})
#define _mm512_cvtps_ph(A, I) __extension__ ({ \ #define _mm512_cvtps_ph(A, I) __extension__ ({ \
(__m256i)__builtin_ia32_vcvtps2ph512_mask((__v16sf)(__m512)(A), (int)(I), \ (__m256i)__builtin_ia32_vcvtps2ph512_mask((__v16sf)(__m512)(A), (int)(I), \
(__v16hi)_mm256_setzero_si256(), \ (__v16hi)_mm256_setzero_si256(), \
(__mmask16)-1); }) (__mmask16)-1); })
#define _mm512_mask_cvtps_ph(U, W, A, I) __extension__ ({ \ #define _mm512_mask_cvtps_ph(U, W, A, I) __extension__ ({ \
(__m256i)__builtin_ia32_vcvtps2ph512_mask((__v16sf)(__m512)(A), (int)(I), \ (__m256i)__builtin_ia32_vcvtps2ph512_mask((__v16sf)(__m512)(A), (int)(I), \
(__v16hi)(__m256i)(U), \ (__v16hi)(__m256i)(U), \
(__mmask16)(W)); }) (__mmask16)(W)); })
#define _mm512_maskz_cvtps_ph(W, A, I) __extension__ ({\ #define _mm512_maskz_cvtps_ph(W, A, I) __extension__ ({\
(__m256i)__builtin_ia32_vcvtps2ph512_mask((__v16sf)(__m512)(A), (int)(I), \ (__m256i)__builtin_ia32_vcvtps2ph512_mask((__v16sf)(__m512)(A), (int)(I), \
(__v16hi)_mm256_setzero_si256(), \ (__v16hi)_mm256_setzero_si256(), \
(__mmask16)(W)); }) (__mmask16)(W)); })
#define _mm512_cvt_roundph_ps( __A, __R) __extension__ ({ \
__builtin_ia32_vcvtph2ps512_mask ((__v16hi)( __A),\
(__v16sf)\
_mm512_undefined_ps (),\
(__mmask16) -1,( __R));\
})
#define _mm512_mask_cvt_roundph_ps( __W, __U, __A, __R) __extension__ ({ \
__builtin_ia32_vcvtph2ps512_mask ((__v16hi)( __A),\
(__v16sf)( __W),\
(__mmask16)( __U),( __R));\
})
#define _mm512_maskz_cvt_roundph_ps( __U, __A, __R) __extension__ ({ \
__builtin_ia32_vcvtph2ps512_mask ((__v16hi)( __A),\
(__v16sf)\
_mm512_setzero_ps (),\
(__mmask16)( __U),( __R));\
})
static __inline __m512 __DEFAULT_FN_ATTRS static __inline __m512 __DEFAULT_FN_ATTRS
_mm512_cvtph_ps(__m256i __A) _mm512_cvtph_ps(__m256i __A)
{ {
return (__m512) __builtin_ia32_vcvtph2ps512_mask ((__v16hi) __A, return (__m512) __builtin_ia32_vcvtph2ps512_mask ((__v16hi) __A,
(__v16sf) (__v16sf)
_mm512_setzero_ps (), _mm512_setzero_ps (),
(__mmask16) -1, (__mmask16) -1,
_MM_FROUND_CUR_DIRECTION); _MM_FROUND_CUR_DIRECTION);
▲ Show 20 Lines • Show All 5,597 Lines • Show Last 20 Lines

test/CodeGen/avx512f-builtins.c

  • This file is larger than 256 KB, so syntax highlighting is disabled by default.
Show First 20 Lines • Show All 3,101 Lines • ▼ Show 20 Lines
} }
unsigned long long test_mm_cvttss_u64(__m128 __A) { unsigned long long test_mm_cvttss_u64(__m128 __A) {
// CHECK-LABEL: @test_mm_cvttss_u64 // CHECK-LABEL: @test_mm_cvttss_u64
// CHECK: @llvm.x86.avx512.cvttss2usi64 // CHECK: @llvm.x86.avx512.cvttss2usi64
return _mm_cvttss_u64(__A); return _mm_cvttss_u64(__A);
} }
__m512i test_mm512_cvtt_roundps_epu32(__m512 __A)
{
// CHECK-LABEL: @test_mm512_cvtt_roundps_epu32
// CHECK: @llvm.x86.avx512.mask.cvttps2udq.512
return _mm512_cvtt_roundps_epu32(__A, _MM_FROUND_CUR_DIRECTION);
}
__m512i test_mm512_mask_cvtt_roundps_epu32(__m512i __W, __mmask16 __U, __m512 __A)
{
// CHECK-LABEL: @test_mm512_mask_cvtt_roundps_epu32
// CHECK: @llvm.x86.avx512.mask.cvttps2udq.512
AsafBadouhUnsubmitted
Not Done
ReplyInline Actions

remove empty line

AsafBadouh: remove empty line
return _mm512_mask_cvtt_roundps_epu32(__W, __U, __A, _MM_FROUND_CUR_DIRECTION);
}
__m512i test_mm512_maskz_cvtt_roundps_epu32( __mmask16 __U, __m512 __A)
{
// CHECK-LABEL: @test_mm512_maskz_cvtt_roundps_epu32
// CHECK: @llvm.x86.avx512.mask.cvttps2udq.512
AsafBadouhUnsubmitted
Not Done
ReplyInline Actions

remove empty line

AsafBadouh: remove empty line
return _mm512_maskz_cvtt_roundps_epu32(__U, __A, _MM_FROUND_CUR_DIRECTION);
}
__m256i test_mm512_cvt_roundps_ph(__m512 __A)
{
// CHECK-LABEL: @test_mm512_cvt_roundps_ph
// CHECK: @llvm.x86.avx512.mask.vcvtps2ph.512
return _mm512_cvt_roundps_ph(__A, _MM_FROUND_CUR_DIRECTION);
}
__m256i test_mm512_mask_cvt_roundps_ph(__m256i __W , __mmask16 __U, __m512 __A)
{
// CHECK-LABEL: @test_mm512_mask_cvt_roundps_ph
// CHECK: @llvm.x86.avx512.mask.vcvtps2ph.512
return _mm512_mask_cvt_roundps_ph(__W, __U, __A, _MM_FROUND_CUR_DIRECTION);
}
__m256i test_mm512_maskz_cvt_roundps_ph(__mmask16 __U, __m512 __A)
{
// CHECK-LABEL: @test_mm512_maskz_cvt_roundps_ph
// CHECK: @llvm.x86.avx512.mask.vcvtps2ph.512
return _mm512_maskz_cvt_roundps_ph(__U, __A, _MM_FROUND_CUR_DIRECTION);
}
__m512 test_mm512_cvt_roundph_ps(__m256i __A)
{
// CHECK-LABEL: @test_mm512_cvt_roundph_ps
// CHECK: @llvm.x86.avx512.mask.vcvtph2ps.512
return _mm512_cvt_roundph_ps(__A, _MM_FROUND_CUR_DIRECTION);
}
__m512 test_mm512_mask_cvt_roundph_ps(__m512 __W, __mmask16 __U, __m256i __A)
{
// CHECK-LABEL: @test_mm512_mask_cvt_roundph_ps
// CHECK: @llvm.x86.avx512.mask.vcvtph2ps.512
return _mm512_mask_cvt_roundph_ps(__W, __U, __A, _MM_FROUND_CUR_DIRECTION);
}
__m512 test_mm512_maskz_cvt_roundph_ps(__mmask16 __U, __m256i __A)
{
// CHECK-LABEL: @test_mm512_maskz_cvt_roundph_ps
// CHECK: @llvm.x86.avx512.mask.vcvtph2ps.512
return _mm512_maskz_cvt_roundph_ps(__U, __A, _MM_FROUND_CUR_DIRECTION);
}
__m512 test_mm512_mask_cvt_roundepi32_ps(__m512 __W, __mmask16 __U, __m512i __A) __m512 test_mm512_mask_cvt_roundepi32_ps(__m512 __W, __mmask16 __U, __m512i __A)
{ {
// CHECK-LABEL: @test_mm512_mask_cvt_roundepi32_ps // CHECK-LABEL: @test_mm512_mask_cvt_roundepi32_ps
// CHECK: @llvm.x86.avx512.mask.cvtdq2ps.512 // CHECK: @llvm.x86.avx512.mask.cvtdq2ps.512
return _mm512_mask_cvt_roundepi32_ps(__W,__U,__A,4); return _mm512_mask_cvt_roundepi32_ps(__W,__U,__A,4);
} }
__m512 test_mm512_maskz_cvt_roundepi32_ps(__mmask16 __U, __m512i __A) __m512 test_mm512_maskz_cvt_roundepi32_ps(__mmask16 __U, __m512i __A)
▲ Show 20 Lines • Show All 3,919 Lines • Show Last 20 Lines