Skip to content

[Headers][X86] Add constexpr support for some AVX[512] intrinsics. #156567

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
RKSimon merged 27 commits into from
Sep 4, 2025
Merged

[Headers][X86] Add constexpr support for some AVX[512] intrinsics. #156567

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
27 commits
Select commit Hold shift + click to select a range
6cdeb1d
_mm_mask_cvtepi32_pd
moorabbit Sep 1, 2025
4438da4
_mm_maskz_cvtepi32_pd
moorabbit Sep 1, 2025
bf013cd
_mm_mask_cvtepi32_ps
moorabbit Sep 1, 2025
034f90d
_mm_maskz_cvtepi32_ps
moorabbit Sep 1, 2025
db7f4a8
_mm_cvtepu32_pd
moorabbit Sep 1, 2025
14ff3b1
_mm_mask_cvtepu32_pd
moorabbit Sep 1, 2025
1c7adb9
_mm_maskz_cvtepu32_pd
moorabbit Sep 1, 2025
2e9e3f2
_mm_cvtepu32_ps
moorabbit Sep 2, 2025
b2aef2b
_mm_mask_cvtepu32_ps
moorabbit Sep 2, 2025
90e7a6b
_mm_maskz_cvtepu32_ps
moorabbit Sep 2, 2025
cec1b65
_mm256_mask_cvtepi32_pd
moorabbit Sep 2, 2025
c9b3f1b
_mm256_maskz_cvtepi32_pd
moorabbit Sep 2, 2025
654d336
_mm256_mask_cvtepi32_ps
moorabbit Sep 2, 2025
33480f4
_mm256_maskz_cvtepi32_ps
moorabbit Sep 2, 2025
a490944
_mm256_cvtepu32_pd
moorabbit Sep 2, 2025
4ecf392
_mm256_mask_cvtepu32_pd
moorabbit Sep 2, 2025
6df80b4
_mm256_maskz_cvtepu32_pd
moorabbit Sep 2, 2025
2d058ae
_mm256_cvtepu32_ps
moorabbit Sep 2, 2025
f21dcae
_mm256_mask_cvtepu32_ps
moorabbit Sep 2, 2025
ac2bb02
_mm256_maskz_cvtepu32_ps
moorabbit Sep 2, 2025
56ef740
_mm512_cvtepi64_pd
moorabbit Sep 2, 2025
a2b5813
_mm512_mask_cvtepi64_pd
moorabbit Sep 2, 2025
a3f8d7b
_mm512_maskz_cvtepi64_pd
moorabbit Sep 2, 2025
8819f7e
_mm512_cvtepu64_pd
moorabbit Sep 2, 2025
c3f802e
_mm512_mask_cvtepu64_pd
moorabbit Sep 2, 2025
c681dbd
_mm512_maskz_cvtepu64_pd
moorabbit Sep 2, 2025
eb9bf71
Merge branch 'main' into constexpr-avx512-intrinsics-part-2
RKSimon Sep 4, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
25 changes: 12 additions & 13 deletions clang/lib/Headers/avx512dqintrin.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -483,21 +483,20 @@ _mm512_maskz_cvtps_epu64 (__mmask8 __U, __m256 __A) {
(__v8di)_mm512_setzero_si512(), \
(__mmask8)(U), (int)(R)))


static __inline__ __m512d __DEFAULT_FN_ATTRS512
_mm512_cvtepi64_pd (__m512i __A) {
static __inline__ __m512d __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_cvtepi64_pd(__m512i __A) {
return (__m512d)__builtin_convertvector((__v8di)__A, __v8df);
}

static __inline__ __m512d __DEFAULT_FN_ATTRS512
_mm512_mask_cvtepi64_pd (__m512d __W, __mmask8 __U, __m512i __A) {
static __inline__ __m512d __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_mask_cvtepi64_pd(__m512d __W, __mmask8 __U, __m512i __A) {
return (__m512d)__builtin_ia32_selectpd_512((__mmask8)__U,
(__v8df)_mm512_cvtepi64_pd(__A),
(__v8df)__W);
}

static __inline__ __m512d __DEFAULT_FN_ATTRS512
_mm512_maskz_cvtepi64_pd (__mmask8 __U, __m512i __A) {
static __inline__ __m512d __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_maskz_cvtepi64_pd(__mmask8 __U, __m512i __A) {
return (__m512d)__builtin_ia32_selectpd_512((__mmask8)__U,
(__v8df)_mm512_cvtepi64_pd(__A),
(__v8df)_mm512_setzero_pd());
Expand Down Expand Up @@ -714,20 +713,20 @@ _mm512_maskz_cvttps_epu64 (__mmask8 __U, __m256 __A) {
(__v8di)_mm512_setzero_si512(), \
(__mmask8)(U), (int)(R)))

static __inline__ __m512d __DEFAULT_FN_ATTRS512
_mm512_cvtepu64_pd (__m512i __A) {
static __inline__ __m512d __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_cvtepu64_pd(__m512i __A) {
return (__m512d)__builtin_convertvector((__v8du)__A, __v8df);
}

static __inline__ __m512d __DEFAULT_FN_ATTRS512
_mm512_mask_cvtepu64_pd (__m512d __W, __mmask8 __U, __m512i __A) {
static __inline__ __m512d __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_mask_cvtepu64_pd(__m512d __W, __mmask8 __U, __m512i __A) {
return (__m512d)__builtin_ia32_selectpd_512((__mmask8)__U,
(__v8df)_mm512_cvtepu64_pd(__A),
(__v8df)__W);
}

static __inline__ __m512d __DEFAULT_FN_ATTRS512
_mm512_maskz_cvtepu64_pd (__mmask8 __U, __m512i __A) {
static __inline__ __m512d __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_maskz_cvtepu64_pd(__mmask8 __U, __m512i __A) {
return (__m512d)__builtin_ia32_selectpd_512((__mmask8)__U,
(__v8df)_mm512_cvtepu64_pd(__A),
(__v8df)_mm512_setzero_pd());
Expand Down
80 changes: 40 additions & 40 deletions clang/lib/Headers/avx512vlintrin.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1732,57 +1732,57 @@ _mm256_mask_compressstoreu_epi32 (void *__P, __mmask8 __U, __m256i __A) {
(__mmask8) __U);
}

static __inline__ __m128d __DEFAULT_FN_ATTRS128
_mm_mask_cvtepi32_pd (__m128d __W, __mmask8 __U, __m128i __A) {
static __inline__ __m128d __DEFAULT_FN_ATTRS128_CONSTEXPR
_mm_mask_cvtepi32_pd(__m128d __W, __mmask8 __U, __m128i __A) {
return (__m128d)__builtin_ia32_selectpd_128((__mmask8) __U,
(__v2df)_mm_cvtepi32_pd(__A),
(__v2df)__W);
}

static __inline__ __m128d __DEFAULT_FN_ATTRS128
_mm_maskz_cvtepi32_pd (__mmask8 __U, __m128i __A) {
static __inline__ __m128d __DEFAULT_FN_ATTRS128_CONSTEXPR
_mm_maskz_cvtepi32_pd(__mmask8 __U, __m128i __A) {
return (__m128d)__builtin_ia32_selectpd_128((__mmask8) __U,
(__v2df)_mm_cvtepi32_pd(__A),
(__v2df)_mm_setzero_pd());
}

static __inline__ __m256d __DEFAULT_FN_ATTRS256
_mm256_mask_cvtepi32_pd (__m256d __W, __mmask8 __U, __m128i __A) {
static __inline__ __m256d __DEFAULT_FN_ATTRS256_CONSTEXPR
_mm256_mask_cvtepi32_pd(__m256d __W, __mmask8 __U, __m128i __A) {
return (__m256d)__builtin_ia32_selectpd_256((__mmask8) __U,
(__v4df)_mm256_cvtepi32_pd(__A),
(__v4df)__W);
}

static __inline__ __m256d __DEFAULT_FN_ATTRS256
_mm256_maskz_cvtepi32_pd (__mmask8 __U, __m128i __A) {
static __inline__ __m256d __DEFAULT_FN_ATTRS256_CONSTEXPR
_mm256_maskz_cvtepi32_pd(__mmask8 __U, __m128i __A) {
return (__m256d)__builtin_ia32_selectpd_256((__mmask8) __U,
(__v4df)_mm256_cvtepi32_pd(__A),
(__v4df)_mm256_setzero_pd());
}

static __inline__ __m128 __DEFAULT_FN_ATTRS128
_mm_mask_cvtepi32_ps (__m128 __W, __mmask8 __U, __m128i __A) {
static __inline__ __m128 __DEFAULT_FN_ATTRS128_CONSTEXPR
_mm_mask_cvtepi32_ps(__m128 __W, __mmask8 __U, __m128i __A) {
return (__m128)__builtin_ia32_selectps_128((__mmask8)__U,
(__v4sf)_mm_cvtepi32_ps(__A),
(__v4sf)__W);
}

static __inline__ __m128 __DEFAULT_FN_ATTRS128
_mm_maskz_cvtepi32_ps (__mmask8 __U, __m128i __A) {
static __inline__ __m128 __DEFAULT_FN_ATTRS128_CONSTEXPR
_mm_maskz_cvtepi32_ps(__mmask8 __U, __m128i __A) {
return (__m128)__builtin_ia32_selectps_128((__mmask8)__U,
(__v4sf)_mm_cvtepi32_ps(__A),
(__v4sf)_mm_setzero_ps());
}

static __inline__ __m256 __DEFAULT_FN_ATTRS256
_mm256_mask_cvtepi32_ps (__m256 __W, __mmask8 __U, __m256i __A) {
static __inline__ __m256 __DEFAULT_FN_ATTRS256_CONSTEXPR
_mm256_mask_cvtepi32_ps(__m256 __W, __mmask8 __U, __m256i __A) {
return (__m256)__builtin_ia32_selectps_256((__mmask8)__U,
(__v8sf)_mm256_cvtepi32_ps(__A),
(__v8sf)__W);
}

static __inline__ __m256 __DEFAULT_FN_ATTRS256
_mm256_maskz_cvtepi32_ps (__mmask8 __U, __m256i __A) {
static __inline__ __m256 __DEFAULT_FN_ATTRS256_CONSTEXPR
_mm256_maskz_cvtepi32_ps(__mmask8 __U, __m256i __A) {
return (__m256)__builtin_ia32_selectps_256((__mmask8)__U,
(__v8sf)_mm256_cvtepi32_ps(__A),
(__v8sf)_mm256_setzero_ps());
Expand Down Expand Up @@ -2143,78 +2143,78 @@ _mm256_maskz_cvttps_epu32 (__mmask8 __U, __m256 __A) {
(__mmask8) __U);
}

static __inline__ __m128d __DEFAULT_FN_ATTRS128
_mm_cvtepu32_pd (__m128i __A) {
static __inline__ __m128d __DEFAULT_FN_ATTRS128_CONSTEXPR
_mm_cvtepu32_pd(__m128i __A) {
return (__m128d) __builtin_convertvector(
__builtin_shufflevector((__v4su)__A, (__v4su)__A, 0, 1), __v2df);
}

static __inline__ __m128d __DEFAULT_FN_ATTRS128
_mm_mask_cvtepu32_pd (__m128d __W, __mmask8 __U, __m128i __A) {
static __inline__ __m128d __DEFAULT_FN_ATTRS128_CONSTEXPR
_mm_mask_cvtepu32_pd(__m128d __W, __mmask8 __U, __m128i __A) {
return (__m128d)__builtin_ia32_selectpd_128((__mmask8) __U,
(__v2df)_mm_cvtepu32_pd(__A),
(__v2df)__W);
}

static __inline__ __m128d __DEFAULT_FN_ATTRS128
_mm_maskz_cvtepu32_pd (__mmask8 __U, __m128i __A) {
static __inline__ __m128d __DEFAULT_FN_ATTRS128_CONSTEXPR
_mm_maskz_cvtepu32_pd(__mmask8 __U, __m128i __A) {
return (__m128d)__builtin_ia32_selectpd_128((__mmask8) __U,
(__v2df)_mm_cvtepu32_pd(__A),
(__v2df)_mm_setzero_pd());
}

static __inline__ __m256d __DEFAULT_FN_ATTRS256
_mm256_cvtepu32_pd (__m128i __A) {
static __inline__ __m256d __DEFAULT_FN_ATTRS256_CONSTEXPR
_mm256_cvtepu32_pd(__m128i __A) {
return (__m256d)__builtin_convertvector((__v4su)__A, __v4df);
}

static __inline__ __m256d __DEFAULT_FN_ATTRS256
_mm256_mask_cvtepu32_pd (__m256d __W, __mmask8 __U, __m128i __A) {
static __inline__ __m256d __DEFAULT_FN_ATTRS256_CONSTEXPR
_mm256_mask_cvtepu32_pd(__m256d __W, __mmask8 __U, __m128i __A) {
return (__m256d)__builtin_ia32_selectpd_256((__mmask8) __U,
(__v4df)_mm256_cvtepu32_pd(__A),
(__v4df)__W);
}

static __inline__ __m256d __DEFAULT_FN_ATTRS256
_mm256_maskz_cvtepu32_pd (__mmask8 __U, __m128i __A) {
static __inline__ __m256d __DEFAULT_FN_ATTRS256_CONSTEXPR
_mm256_maskz_cvtepu32_pd(__mmask8 __U, __m128i __A) {
return (__m256d)__builtin_ia32_selectpd_256((__mmask8) __U,
(__v4df)_mm256_cvtepu32_pd(__A),
(__v4df)_mm256_setzero_pd());
}

static __inline__ __m128 __DEFAULT_FN_ATTRS128
_mm_cvtepu32_ps (__m128i __A) {
static __inline__ __m128 __DEFAULT_FN_ATTRS128_CONSTEXPR
_mm_cvtepu32_ps(__m128i __A) {
return (__m128)__builtin_convertvector((__v4su)__A, __v4sf);
}

static __inline__ __m128 __DEFAULT_FN_ATTRS128
_mm_mask_cvtepu32_ps (__m128 __W, __mmask8 __U, __m128i __A) {
static __inline__ __m128 __DEFAULT_FN_ATTRS128_CONSTEXPR
_mm_mask_cvtepu32_ps(__m128 __W, __mmask8 __U, __m128i __A) {
return (__m128)__builtin_ia32_selectps_128((__mmask8)__U,
(__v4sf)_mm_cvtepu32_ps(__A),
(__v4sf)__W);
}

static __inline__ __m128 __DEFAULT_FN_ATTRS128
_mm_maskz_cvtepu32_ps (__mmask8 __U, __m128i __A) {
static __inline__ __m128 __DEFAULT_FN_ATTRS128_CONSTEXPR
_mm_maskz_cvtepu32_ps(__mmask8 __U, __m128i __A) {
return (__m128)__builtin_ia32_selectps_128((__mmask8)__U,
(__v4sf)_mm_cvtepu32_ps(__A),
(__v4sf)_mm_setzero_ps());
}

static __inline__ __m256 __DEFAULT_FN_ATTRS256
_mm256_cvtepu32_ps (__m256i __A) {
static __inline__ __m256 __DEFAULT_FN_ATTRS256_CONSTEXPR
_mm256_cvtepu32_ps(__m256i __A) {
return (__m256)__builtin_convertvector((__v8su)__A, __v8sf);
}

static __inline__ __m256 __DEFAULT_FN_ATTRS256
_mm256_mask_cvtepu32_ps (__m256 __W, __mmask8 __U, __m256i __A) {
static __inline__ __m256 __DEFAULT_FN_ATTRS256_CONSTEXPR
_mm256_mask_cvtepu32_ps(__m256 __W, __mmask8 __U, __m256i __A) {
return (__m256)__builtin_ia32_selectps_256((__mmask8)__U,
(__v8sf)_mm256_cvtepu32_ps(__A),
(__v8sf)__W);
}

static __inline__ __m256 __DEFAULT_FN_ATTRS256
_mm256_maskz_cvtepu32_ps (__mmask8 __U, __m256i __A) {
static __inline__ __m256 __DEFAULT_FN_ATTRS256_CONSTEXPR
_mm256_maskz_cvtepu32_ps(__mmask8 __U, __m256i __A) {
return (__m256)__builtin_ia32_selectps_256((__mmask8)__U,
(__v8sf)_mm256_cvtepu32_ps(__A),
(__v8sf)_mm256_setzero_ps());
Expand Down
12 changes: 12 additions & 0 deletions clang/test/CodeGen/X86/avx512dq-builtins.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -602,20 +602,26 @@ __m512d test_mm512_cvtepi64_pd(__m512i __A) {
return _mm512_cvtepi64_pd(__A);
}

TEST_CONSTEXPR(match_m512d(_mm512_cvtepi64_pd((__m512i)(__v8di){-1, -1, 2, 2, -4, -4, 8, 8}), -1.0, -1.0, 2.0, 2.0, -4.0, -4.0, 8.0, 8.0));

__m512d test_mm512_mask_cvtepi64_pd(__m512d __W, __mmask8 __U, __m512i __A) {
// CHECK-LABEL: test_mm512_mask_cvtepi64_pd
// CHECK: sitofp <8 x i64> %{{.*}} to <8 x double>
// CHECK: select <8 x i1> %{{.*}}, <8 x double> %{{.*}}, <8 x double> %{{.*}}
return _mm512_mask_cvtepi64_pd(__W, __U, __A);
}

TEST_CONSTEXPR(match_m512d(_mm512_mask_cvtepi64_pd(_mm512_set1_pd(-777.0), /*1010 0101=*/0xa5, (__m512i)(__v8di){-1, -1, 2, 2, -4, -4, 8, 8}), -1.0, -777.0, 2.0, -777.0, -777.0, -4.0, -777.0, 8.0));

__m512d test_mm512_maskz_cvtepi64_pd(__mmask8 __U, __m512i __A) {
// CHECK-LABEL: test_mm512_maskz_cvtepi64_pd
// CHECK: sitofp <8 x i64> %{{.*}} to <8 x double>
// CHECK: select <8 x i1> %{{.*}}, <8 x double> %{{.*}}, <8 x double> %{{.*}}
return _mm512_maskz_cvtepi64_pd(__U, __A);
}

TEST_CONSTEXPR(match_m512d(_mm512_maskz_cvtepi64_pd(/*1010 0101=*/0xa5, (__m512i)(__v8di){-1, -1, 2, 2, -4, -4, 8, 8}), -1.0, 0.0, 2.0, 0.0, 0.0, -4.0, 0.0, 8.0));

__m512d test_mm512_cvt_roundepi64_pd(__m512i __A) {
// CHECK-LABEL: test_mm512_cvt_roundepi64_pd
// CHECK: @llvm.x86.avx512.sitofp.round.v8f64.v8i64
Expand Down Expand Up @@ -826,20 +832,26 @@ __m512d test_mm512_cvtepu64_pd(__m512i __A) {
return _mm512_cvtepu64_pd(__A);
}

TEST_CONSTEXPR(match_m512d(_mm512_cvtepu64_pd((__m512i)(__v8du){1, 1, 2, 2, 4, 4, 8, 8}), 1.0, 1.0, 2.0, 2.0, 4.0, 4.0, 8.0, 8.0));

__m512d test_mm512_mask_cvtepu64_pd(__m512d __W, __mmask8 __U, __m512i __A) {
// CHECK-LABEL: test_mm512_mask_cvtepu64_pd
// CHECK: uitofp <8 x i64> %{{.*}} to <8 x double>
// CHECK: select <8 x i1> %{{.*}}, <8 x double> %{{.*}}, <8 x double> %{{.*}}
return _mm512_mask_cvtepu64_pd(__W, __U, __A);
}

TEST_CONSTEXPR(match_m512d(_mm512_mask_cvtepu64_pd(_mm512_set1_pd(-777.0), /*1010 0101=*/0xa5, (__m512i)(__v8du){1, 1, 2, 2, 4, 4, 8, 8}), 1.0, -777.0, 2.0, -777.0, -777.0, 4.0, -777.0, 8.0));

__m512d test_mm512_maskz_cvtepu64_pd(__mmask8 __U, __m512i __A) {
// CHECK-LABEL: test_mm512_maskz_cvtepu64_pd
// CHECK: uitofp <8 x i64> %{{.*}} to <8 x double>
// CHECK: select <8 x i1> %{{.*}}, <8 x double> %{{.*}}, <8 x double> %{{.*}}
return _mm512_maskz_cvtepu64_pd(__U, __A);
}

TEST_CONSTEXPR(match_m512d(_mm512_maskz_cvtepu64_pd(/*1010 0101=*/0xa5, (__m512i)(__v8du){1, 1, 2, 2, 4, 4, 8, 8}), 1.0, 0.0, 2.0, 0.0, 0.0, 4.0, 0.0, 8.0));

__m512d test_mm512_cvt_roundepu64_pd(__m512i __A) {
// CHECK-LABEL: test_mm512_cvt_roundepu64_pd
// CHECK: @llvm.x86.avx512.uitofp.round.v8f64.v8i64
Expand Down
Loading
Loading