Skip to content

[X86] Allow AVX512 funnel shift intrinsics to be used in constexpr #157668

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 1 commit into from
Sep 9, 2025
Merged

[X86] Allow AVX512 funnel shift intrinsics to be used in constexpr #157668

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
43 changes: 25 additions & 18 deletions clang/lib/Headers/avx512vbmi2intrin.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,6 +19,12 @@
__attribute__((__always_inline__, __nodebug__, __target__("avx512vbmi2"), \
__min_vector_width__(512)))

#if defined(__cplusplus) && (__cplusplus >= 201103L)
#define __DEFAULT_FN_ATTRS_CONSTEXPR __DEFAULT_FN_ATTRS constexpr
#else
#define __DEFAULT_FN_ATTRS_CONSTEXPR __DEFAULT_FN_ATTRS
#endif

static __inline__ __m512i __DEFAULT_FN_ATTRS
_mm512_mask_compress_epi16(__m512i __S, __mmask32 __U, __m512i __D)
{
Expand Down Expand Up @@ -213,140 +219,140 @@ _mm512_maskz_expandloadu_epi8(__mmask64 __U, void const *__P)
(__v32hi)_mm512_shrdi_epi16((A), (B), (I)), \
(__v32hi)_mm512_setzero_si512()))

static __inline__ __m512i __DEFAULT_FN_ATTRS
static __inline__ __m512i __DEFAULT_FN_ATTRS_CONSTEXPR
_mm512_shldv_epi64(__m512i __A, __m512i __B, __m512i __C)
{
return (__m512i)__builtin_elementwise_fshl((__v8du)__A, (__v8du)__B,
(__v8du)__C);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS
static __inline__ __m512i __DEFAULT_FN_ATTRS_CONSTEXPR
_mm512_mask_shldv_epi64(__m512i __A, __mmask8 __U, __m512i __B, __m512i __C)
{
return (__m512i)__builtin_ia32_selectq_512(__U,
(__v8di)_mm512_shldv_epi64(__A, __B, __C),
(__v8di)__A);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS
static __inline__ __m512i __DEFAULT_FN_ATTRS_CONSTEXPR
_mm512_maskz_shldv_epi64(__mmask8 __U, __m512i __A, __m512i __B, __m512i __C)
{
return (__m512i)__builtin_ia32_selectq_512(__U,
(__v8di)_mm512_shldv_epi64(__A, __B, __C),
(__v8di)_mm512_setzero_si512());
}

static __inline__ __m512i __DEFAULT_FN_ATTRS
static __inline__ __m512i __DEFAULT_FN_ATTRS_CONSTEXPR
_mm512_shldv_epi32(__m512i __A, __m512i __B, __m512i __C)
{
return (__m512i)__builtin_elementwise_fshl((__v16su)__A, (__v16su)__B,
(__v16su)__C);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS
static __inline__ __m512i __DEFAULT_FN_ATTRS_CONSTEXPR
_mm512_mask_shldv_epi32(__m512i __A, __mmask16 __U, __m512i __B, __m512i __C)
{
return (__m512i)__builtin_ia32_selectd_512(__U,
(__v16si)_mm512_shldv_epi32(__A, __B, __C),
(__v16si)__A);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS
static __inline__ __m512i __DEFAULT_FN_ATTRS_CONSTEXPR
_mm512_maskz_shldv_epi32(__mmask16 __U, __m512i __A, __m512i __B, __m512i __C)
{
return (__m512i)__builtin_ia32_selectd_512(__U,
(__v16si)_mm512_shldv_epi32(__A, __B, __C),
(__v16si)_mm512_setzero_si512());
}

static __inline__ __m512i __DEFAULT_FN_ATTRS
static __inline__ __m512i __DEFAULT_FN_ATTRS_CONSTEXPR
_mm512_shldv_epi16(__m512i __A, __m512i __B, __m512i __C)
{
return (__m512i)__builtin_elementwise_fshl((__v32hu)__A, (__v32hu)__B,
(__v32hu)__C);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS
static __inline__ __m512i __DEFAULT_FN_ATTRS_CONSTEXPR
_mm512_mask_shldv_epi16(__m512i __A, __mmask32 __U, __m512i __B, __m512i __C)
{
return (__m512i)__builtin_ia32_selectw_512(__U,
(__v32hi)_mm512_shldv_epi16(__A, __B, __C),
(__v32hi)__A);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS
static __inline__ __m512i __DEFAULT_FN_ATTRS_CONSTEXPR
_mm512_maskz_shldv_epi16(__mmask32 __U, __m512i __A, __m512i __B, __m512i __C)
{
return (__m512i)__builtin_ia32_selectw_512(__U,
(__v32hi)_mm512_shldv_epi16(__A, __B, __C),
(__v32hi)_mm512_setzero_si512());
}

static __inline__ __m512i __DEFAULT_FN_ATTRS
static __inline__ __m512i __DEFAULT_FN_ATTRS_CONSTEXPR
_mm512_shrdv_epi64(__m512i __A, __m512i __B, __m512i __C)
{
// Ops __A and __B are swapped.
return (__m512i)__builtin_elementwise_fshr((__v8du)__B, (__v8du)__A,
(__v8du)__C);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS
static __inline__ __m512i __DEFAULT_FN_ATTRS_CONSTEXPR
_mm512_mask_shrdv_epi64(__m512i __A, __mmask8 __U, __m512i __B, __m512i __C)
{
return (__m512i)__builtin_ia32_selectq_512(__U,
(__v8di)_mm512_shrdv_epi64(__A, __B, __C),
(__v8di)__A);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS
static __inline__ __m512i __DEFAULT_FN_ATTRS_CONSTEXPR
_mm512_maskz_shrdv_epi64(__mmask8 __U, __m512i __A, __m512i __B, __m512i __C)
{
return (__m512i)__builtin_ia32_selectq_512(__U,
(__v8di)_mm512_shrdv_epi64(__A, __B, __C),
(__v8di)_mm512_setzero_si512());
}

static __inline__ __m512i __DEFAULT_FN_ATTRS
static __inline__ __m512i __DEFAULT_FN_ATTRS_CONSTEXPR
_mm512_shrdv_epi32(__m512i __A, __m512i __B, __m512i __C)
{
// Ops __A and __B are swapped.
return (__m512i)__builtin_elementwise_fshr((__v16su)__B, (__v16su)__A,
(__v16su)__C);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS
static __inline__ __m512i __DEFAULT_FN_ATTRS_CONSTEXPR
_mm512_mask_shrdv_epi32(__m512i __A, __mmask16 __U, __m512i __B, __m512i __C)
{
return (__m512i) __builtin_ia32_selectd_512(__U,
(__v16si)_mm512_shrdv_epi32(__A, __B, __C),
(__v16si)__A);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS
static __inline__ __m512i __DEFAULT_FN_ATTRS_CONSTEXPR
_mm512_maskz_shrdv_epi32(__mmask16 __U, __m512i __A, __m512i __B, __m512i __C)
{
return (__m512i) __builtin_ia32_selectd_512(__U,
(__v16si)_mm512_shrdv_epi32(__A, __B, __C),
(__v16si)_mm512_setzero_si512());
}

static __inline__ __m512i __DEFAULT_FN_ATTRS
static __inline__ __m512i __DEFAULT_FN_ATTRS_CONSTEXPR
_mm512_shrdv_epi16(__m512i __A, __m512i __B, __m512i __C)
{
// Ops __A and __B are swapped.
return (__m512i)__builtin_elementwise_fshr((__v32hu)__B, (__v32hu)__A,
(__v32hu)__C);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS
static __inline__ __m512i __DEFAULT_FN_ATTRS_CONSTEXPR
_mm512_mask_shrdv_epi16(__m512i __A, __mmask32 __U, __m512i __B, __m512i __C)
{
return (__m512i)__builtin_ia32_selectw_512(__U,
(__v32hi)_mm512_shrdv_epi16(__A, __B, __C),
(__v32hi)__A);
}

static __inline__ __m512i __DEFAULT_FN_ATTRS
static __inline__ __m512i __DEFAULT_FN_ATTRS_CONSTEXPR
_mm512_maskz_shrdv_epi16(__mmask32 __U, __m512i __A, __m512i __B, __m512i __C)
{
return (__m512i)__builtin_ia32_selectw_512(__U,
Expand All @@ -356,6 +362,7 @@ _mm512_maskz_shrdv_epi16(__mmask32 __U, __m512i __A, __m512i __B, __m512i __C)


#undef __DEFAULT_FN_ATTRS
#undef __DEFAULT_FN_ATTRS_CONSTEXPR

#endif

Loading
Loading