Skip to content

[Headers][X86] Allow AVX512 integer min/max mask/maskz variants intrinsics to be used in constexpr #156901

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
merged 3 commits into from
Sep 6, 2025
Merged

[Headers][X86] Allow AVX512 integer min/max mask/maskz variants intrinsics to be used in constexpr #156901

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
2b91d9f
[Headers][X86] Allow AVX512 integer min/max mask/maskz variants intri…
markbhasawut Sep 4, 2025
ab99aea
Update mask/maskz variants in avx512vlintrin.h and avx512vlbwintrin.h to
markbhasawut Sep 5, 2025
9c20815
Merge branch 'main' into x86-avx512-mask-constexpr-intrinsics
RKSimon Sep 6, 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
81 changes: 32 additions & 49 deletions clang/lib/Headers/avx512bwintrin.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -741,17 +741,15 @@ _mm512_max_epi8(__m512i __A, __m512i __B) {
return (__m512i)__builtin_elementwise_max((__v64qs) __A, (__v64qs) __B);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS512
_mm512_maskz_max_epi8 (__mmask64 __M, __m512i __A, __m512i __B)
{
static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_maskz_max_epi8(__mmask64 __M, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectb_512((__mmask64)__M,
(__v64qi)_mm512_max_epi8(__A, __B),
(__v64qi)_mm512_setzero_si512());
}

static __inline__ __m512i __DEFAULT_FN_ATTRS512
_mm512_mask_max_epi8 (__m512i __W, __mmask64 __M, __m512i __A, __m512i __B)
{
static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_mask_max_epi8(__m512i __W, __mmask64 __M, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectb_512((__mmask64)__M,
(__v64qi)_mm512_max_epi8(__A, __B),
(__v64qi)__W);
Expand All @@ -762,18 +760,15 @@ _mm512_max_epi16(__m512i __A, __m512i __B) {
return (__m512i)__builtin_elementwise_max((__v32hi) __A, (__v32hi) __B);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS512
_mm512_maskz_max_epi16 (__mmask32 __M, __m512i __A, __m512i __B)
{
static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_maskz_max_epi16(__mmask32 __M, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectw_512((__mmask32)__M,
(__v32hi)_mm512_max_epi16(__A, __B),
(__v32hi)_mm512_setzero_si512());
}

static __inline__ __m512i __DEFAULT_FN_ATTRS512
_mm512_mask_max_epi16 (__m512i __W, __mmask32 __M, __m512i __A,
__m512i __B)
{
static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_mask_max_epi16(__m512i __W, __mmask32 __M, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectw_512((__mmask32)__M,
(__v32hi)_mm512_max_epi16(__A, __B),
(__v32hi)__W);
Expand All @@ -784,17 +779,15 @@ _mm512_max_epu8(__m512i __A, __m512i __B) {
return (__m512i)__builtin_elementwise_max((__v64qu)__A, (__v64qu)__B);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS512
_mm512_maskz_max_epu8 (__mmask64 __M, __m512i __A, __m512i __B)
{
static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_maskz_max_epu8(__mmask64 __M, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectb_512((__mmask64)__M,
(__v64qi)_mm512_max_epu8(__A, __B),
(__v64qi)_mm512_setzero_si512());
}

static __inline__ __m512i __DEFAULT_FN_ATTRS512
_mm512_mask_max_epu8 (__m512i __W, __mmask64 __M, __m512i __A, __m512i __B)
{
static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_mask_max_epu8(__m512i __W, __mmask64 __M, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectb_512((__mmask64)__M,
(__v64qi)_mm512_max_epu8(__A, __B),
(__v64qi)__W);
Expand All @@ -805,17 +798,15 @@ _mm512_max_epu16(__m512i __A, __m512i __B) {
return (__m512i)__builtin_elementwise_max((__v32hu)__A, (__v32hu)__B);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS512
_mm512_maskz_max_epu16 (__mmask32 __M, __m512i __A, __m512i __B)
{
static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_maskz_max_epu16(__mmask32 __M, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectw_512((__mmask32)__M,
(__v32hi)_mm512_max_epu16(__A, __B),
(__v32hi)_mm512_setzero_si512());
}

static __inline__ __m512i __DEFAULT_FN_ATTRS512
_mm512_mask_max_epu16 (__m512i __W, __mmask32 __M, __m512i __A, __m512i __B)
{
static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_mask_max_epu16(__m512i __W, __mmask32 __M, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectw_512((__mmask32)__M,
(__v32hi)_mm512_max_epu16(__A, __B),
(__v32hi)__W);
Expand All @@ -826,17 +817,15 @@ _mm512_min_epi8(__m512i __A, __m512i __B) {
return (__m512i)__builtin_elementwise_min((__v64qs) __A, (__v64qs) __B);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS512
_mm512_maskz_min_epi8 (__mmask64 __M, __m512i __A, __m512i __B)
{
static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_maskz_min_epi8(__mmask64 __M, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectb_512((__mmask64)__M,
(__v64qi)_mm512_min_epi8(__A, __B),
(__v64qi)_mm512_setzero_si512());
}

static __inline__ __m512i __DEFAULT_FN_ATTRS512
_mm512_mask_min_epi8 (__m512i __W, __mmask64 __M, __m512i __A, __m512i __B)
{
static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_mask_min_epi8(__m512i __W, __mmask64 __M, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectb_512((__mmask64)__M,
(__v64qi)_mm512_min_epi8(__A, __B),
(__v64qi)__W);
Expand All @@ -847,17 +836,15 @@ _mm512_min_epi16(__m512i __A, __m512i __B) {
return (__m512i)__builtin_elementwise_min((__v32hi) __A, (__v32hi) __B);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS512
_mm512_maskz_min_epi16 (__mmask32 __M, __m512i __A, __m512i __B)
{
static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_maskz_min_epi16(__mmask32 __M, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectw_512((__mmask32)__M,
(__v32hi)_mm512_min_epi16(__A, __B),
(__v32hi)_mm512_setzero_si512());
}

static __inline__ __m512i __DEFAULT_FN_ATTRS512
_mm512_mask_min_epi16 (__m512i __W, __mmask32 __M, __m512i __A, __m512i __B)
{
static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_mask_min_epi16(__m512i __W, __mmask32 __M, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectw_512((__mmask32)__M,
(__v32hi)_mm512_min_epi16(__A, __B),
(__v32hi)__W);
Expand All @@ -868,17 +855,15 @@ _mm512_min_epu8(__m512i __A, __m512i __B) {
return (__m512i)__builtin_elementwise_min((__v64qu)__A, (__v64qu)__B);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS512
_mm512_maskz_min_epu8 (__mmask64 __M, __m512i __A, __m512i __B)
{
static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_maskz_min_epu8(__mmask64 __M, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectb_512((__mmask64)__M,
(__v64qi)_mm512_min_epu8(__A, __B),
(__v64qi)_mm512_setzero_si512());
}

static __inline__ __m512i __DEFAULT_FN_ATTRS512
_mm512_mask_min_epu8 (__m512i __W, __mmask64 __M, __m512i __A, __m512i __B)
{
static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_mask_min_epu8(__m512i __W, __mmask64 __M, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectb_512((__mmask64)__M,
(__v64qi)_mm512_min_epu8(__A, __B),
(__v64qi)__W);
Expand All @@ -889,17 +874,15 @@ _mm512_min_epu16(__m512i __A, __m512i __B) {
return (__m512i)__builtin_elementwise_min((__v32hu)__A, (__v32hu)__B);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS512
_mm512_maskz_min_epu16 (__mmask32 __M, __m512i __A, __m512i __B)
{
static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_maskz_min_epu16(__mmask32 __M, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectw_512((__mmask32)__M,
(__v32hi)_mm512_min_epu16(__A, __B),
(__v32hi)_mm512_setzero_si512());
}

static __inline__ __m512i __DEFAULT_FN_ATTRS512
_mm512_mask_min_epu16 (__m512i __W, __mmask32 __M, __m512i __A, __m512i __B)
{
static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_mask_min_epu16(__m512i __W, __mmask32 __M, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectw_512((__mmask32)__M,
(__v32hi)_mm512_min_epu16(__A, __B),
(__v32hi)__W);
Expand Down
80 changes: 32 additions & 48 deletions clang/lib/Headers/avx512fintrin.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1092,17 +1092,15 @@ static __inline __m512i
return (__m512i)__builtin_elementwise_max((__v16si)__A, (__v16si)__B);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS512
_mm512_mask_max_epi32 (__m512i __W, __mmask16 __M, __m512i __A, __m512i __B)
{
static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_mask_max_epi32(__m512i __W, __mmask16 __M, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectd_512((__mmask16)__M,
(__v16si)_mm512_max_epi32(__A, __B),
(__v16si)__W);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS512
_mm512_maskz_max_epi32 (__mmask16 __M, __m512i __A, __m512i __B)
{
static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_maskz_max_epi32(__mmask16 __M, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectd_512((__mmask16)__M,
(__v16si)_mm512_max_epi32(__A, __B),
(__v16si)_mm512_setzero_si512());
Expand All @@ -1113,17 +1111,15 @@ _mm512_max_epu32(__m512i __A, __m512i __B) {
return (__m512i)__builtin_elementwise_max((__v16su)__A, (__v16su)__B);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS512
_mm512_mask_max_epu32 (__m512i __W, __mmask16 __M, __m512i __A, __m512i __B)
{
static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_mask_max_epu32(__m512i __W, __mmask16 __M, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectd_512((__mmask16)__M,
(__v16si)_mm512_max_epu32(__A, __B),
(__v16si)__W);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS512
_mm512_maskz_max_epu32 (__mmask16 __M, __m512i __A, __m512i __B)
{
static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_maskz_max_epu32(__mmask16 __M, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectd_512((__mmask16)__M,
(__v16si)_mm512_max_epu32(__A, __B),
(__v16si)_mm512_setzero_si512());
Expand All @@ -1134,17 +1130,15 @@ _mm512_max_epi64(__m512i __A, __m512i __B) {
return (__m512i)__builtin_elementwise_max((__v8di)__A, (__v8di)__B);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS512
_mm512_mask_max_epi64 (__m512i __W, __mmask8 __M, __m512i __A, __m512i __B)
{
static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_mask_max_epi64(__m512i __W, __mmask8 __M, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectq_512((__mmask8)__M,
(__v8di)_mm512_max_epi64(__A, __B),
(__v8di)__W);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS512
_mm512_maskz_max_epi64 (__mmask8 __M, __m512i __A, __m512i __B)
{
static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_maskz_max_epi64(__mmask8 __M, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectq_512((__mmask8)__M,
(__v8di)_mm512_max_epi64(__A, __B),
(__v8di)_mm512_setzero_si512());
Expand All @@ -1155,17 +1149,15 @@ _mm512_max_epu64(__m512i __A, __m512i __B) {
return (__m512i)__builtin_elementwise_max((__v8du)__A, (__v8du)__B);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS512
_mm512_mask_max_epu64 (__m512i __W, __mmask8 __M, __m512i __A, __m512i __B)
{
static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_mask_max_epu64(__m512i __W, __mmask8 __M, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectq_512((__mmask8)__M,
(__v8di)_mm512_max_epu64(__A, __B),
(__v8di)__W);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS512
_mm512_maskz_max_epu64 (__mmask8 __M, __m512i __A, __m512i __B)
{
static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_maskz_max_epu64(__mmask8 __M, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectq_512((__mmask8)__M,
(__v8di)_mm512_max_epu64(__A, __B),
(__v8di)_mm512_setzero_si512());
Expand Down Expand Up @@ -1322,17 +1314,15 @@ static __inline __m512i
return (__m512i)__builtin_elementwise_min((__v16si)__A, (__v16si)__B);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS512
_mm512_mask_min_epi32 (__m512i __W, __mmask16 __M, __m512i __A, __m512i __B)
{
static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_mask_min_epi32(__m512i __W, __mmask16 __M, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectd_512((__mmask16)__M,
(__v16si)_mm512_min_epi32(__A, __B),
(__v16si)__W);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS512
_mm512_maskz_min_epi32 (__mmask16 __M, __m512i __A, __m512i __B)
{
static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_maskz_min_epi32(__mmask16 __M, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectd_512((__mmask16)__M,
(__v16si)_mm512_min_epi32(__A, __B),
(__v16si)_mm512_setzero_si512());
Expand All @@ -1343,17 +1333,15 @@ _mm512_min_epu32(__m512i __A, __m512i __B) {
return (__m512i)__builtin_elementwise_min((__v16su)__A, (__v16su)__B);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS512
_mm512_mask_min_epu32 (__m512i __W, __mmask16 __M, __m512i __A, __m512i __B)
{
static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_mask_min_epu32(__m512i __W, __mmask16 __M, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectd_512((__mmask16)__M,
(__v16si)_mm512_min_epu32(__A, __B),
(__v16si)__W);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS512
_mm512_maskz_min_epu32 (__mmask16 __M, __m512i __A, __m512i __B)
{
static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_maskz_min_epu32(__mmask16 __M, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectd_512((__mmask16)__M,
(__v16si)_mm512_min_epu32(__A, __B),
(__v16si)_mm512_setzero_si512());
Expand All @@ -1364,17 +1352,15 @@ _mm512_min_epi64(__m512i __A, __m512i __B) {
return (__m512i)__builtin_elementwise_min((__v8di)__A, (__v8di)__B);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS512
_mm512_mask_min_epi64 (__m512i __W, __mmask8 __M, __m512i __A, __m512i __B)
{
static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_mask_min_epi64(__m512i __W, __mmask8 __M, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectq_512((__mmask8)__M,
(__v8di)_mm512_min_epi64(__A, __B),
(__v8di)__W);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS512
_mm512_maskz_min_epi64 (__mmask8 __M, __m512i __A, __m512i __B)
{
static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_maskz_min_epi64(__mmask8 __M, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectq_512((__mmask8)__M,
(__v8di)_mm512_min_epi64(__A, __B),
(__v8di)_mm512_setzero_si512());
Expand All @@ -1385,17 +1371,15 @@ _mm512_min_epu64(__m512i __A, __m512i __B) {
return (__m512i)__builtin_elementwise_min((__v8du)__A, (__v8du)__B);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS512
_mm512_mask_min_epu64 (__m512i __W, __mmask8 __M, __m512i __A, __m512i __B)
{
static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_mask_min_epu64(__m512i __W, __mmask8 __M, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectq_512((__mmask8)__M,
(__v8di)_mm512_min_epu64(__A, __B),
(__v8di)__W);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS512
_mm512_maskz_min_epu64 (__mmask8 __M, __m512i __A, __m512i __B)
{
static __inline__ __m512i __DEFAULT_FN_ATTRS512_CONSTEXPR
_mm512_maskz_min_epu64(__mmask8 __M, __m512i __A, __m512i __B) {
return (__m512i)__builtin_ia32_selectq_512((__mmask8)__M,
(__v8di)_mm512_min_epu64(__A, __B),
(__v8di)_mm512_setzero_si512());
Expand Down
Loading