Mirrors/zig
1
0
Fork 0
mirror of https://codeberg.org/ziglang/zig.git synced 2026-04-26 13:01:34 +03:00
Code Issues Packages Projects Releases Wiki Activity
Files
master
zig/lib/include/fma4intrin.h
T
Alex Rønne Petersen 520af69660 zig cc: update intrinsic headers to LLVM 22
2026-04-25 21:49:39 +02:00

221 lines
8.1 KiB
C++
Vendored
Raw Permalink Blame History

/*===---- fma4intrin.h - FMA4 intrinsics -----------------------------------===
*
* Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
* See https://llvm.org/LICENSE.txt for license information.
* SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
*
*===-----------------------------------------------------------------------===
*/
#ifndef __X86INTRIN_H
#error "Never use <fma4intrin.h> directly; include <x86intrin.h> instead."
#endif
#ifndef __FMA4INTRIN_H
#define __FMA4INTRIN_H
#include <pmmintrin.h>
/* Define the default attributes for the functions in this file. */
#define __DEFAULT_FN_ATTRS128 __attribute__((__always_inline__, __nodebug__, __target__("fma4"), __min_vector_width__(128)))
#define __DEFAULT_FN_ATTRS256 __attribute__((__always_inline__, __nodebug__, __target__("fma4"), __min_vector_width__(256)))
#if defined(__cplusplus) && (__cplusplus >= 201103L)
#define __DEFAULT_FN_ATTRS128_CONSTEXPR __DEFAULT_FN_ATTRS128 constexpr
#define __DEFAULT_FN_ATTRS256_CONSTEXPR __DEFAULT_FN_ATTRS256 constexpr
#else
#define __DEFAULT_FN_ATTRS128_CONSTEXPR __DEFAULT_FN_ATTRS128
#define __DEFAULT_FN_ATTRS256_CONSTEXPR __DEFAULT_FN_ATTRS256
#endif
static __inline__ __m128 __DEFAULT_FN_ATTRS128_CONSTEXPR
_mm_macc_ps(__m128 __A, __m128 __B, __m128 __C) {
return (__m128)__builtin_elementwise_fma((__v4sf)__A, (__v4sf)__B,
(__v4sf)__C);
}
static __inline__ __m128d __DEFAULT_FN_ATTRS128_CONSTEXPR
_mm_macc_pd(__m128d __A, __m128d __B, __m128d __C) {
return (__m128d)__builtin_elementwise_fma((__v2df)__A, (__v2df)__B,
(__v2df)__C);
}
static __inline__ __m128 __DEFAULT_FN_ATTRS128_CONSTEXPR
_mm_macc_ss(__m128 __A, __m128 __B, __m128 __C) {
return _mm_set_ss(__builtin_elementwise_fma(__A[0], __B[0], __C[0]));
}
static __inline__ __m128d __DEFAULT_FN_ATTRS128_CONSTEXPR
_mm_macc_sd(__m128d __A, __m128d __B, __m128d __C) {
return _mm_set_sd(__builtin_elementwise_fma(__A[0], __B[0], __C[0]));
}
static __inline__ __m128 __DEFAULT_FN_ATTRS128_CONSTEXPR
_mm_msub_ps(__m128 __A, __m128 __B, __m128 __C) {
return (__m128)__builtin_elementwise_fma((__v4sf)__A, (__v4sf)__B,
-(__v4sf)__C);
}
static __inline__ __m128d __DEFAULT_FN_ATTRS128_CONSTEXPR
_mm_msub_pd(__m128d __A, __m128d __B, __m128d __C) {
return (__m128d)__builtin_elementwise_fma((__v2df)__A, (__v2df)__B,
-(__v2df)__C);
}
static __inline__ __m128 __DEFAULT_FN_ATTRS128_CONSTEXPR
_mm_msub_ss(__m128 __A, __m128 __B, __m128 __C) {
return _mm_set_ss(__builtin_elementwise_fma(__A[0], __B[0], -__C[0]));
}
static __inline__ __m128d __DEFAULT_FN_ATTRS128_CONSTEXPR
_mm_msub_sd(__m128d __A, __m128d __B, __m128d __C) {
return _mm_set_sd(__builtin_elementwise_fma(__A[0], __B[0], -__C[0]));
}
static __inline__ __m128 __DEFAULT_FN_ATTRS128_CONSTEXPR
_mm_nmacc_ps(__m128 __A, __m128 __B, __m128 __C) {
return (__m128)__builtin_elementwise_fma(-(__v4sf)__A, (__v4sf)__B,
(__v4sf)__C);
}
static __inline__ __m128d __DEFAULT_FN_ATTRS128_CONSTEXPR
_mm_nmacc_pd(__m128d __A, __m128d __B, __m128d __C) {
return (__m128d)__builtin_elementwise_fma(-(__v2df)__A, (__v2df)__B,
(__v2df)__C);
}
static __inline__ __m128 __DEFAULT_FN_ATTRS128_CONSTEXPR
_mm_nmacc_ss(__m128 __A, __m128 __B, __m128 __C) {
return _mm_set_ss(__builtin_elementwise_fma(-__A[0], __B[0], __C[0]));
}
static __inline__ __m128d __DEFAULT_FN_ATTRS128_CONSTEXPR
_mm_nmacc_sd(__m128d __A, __m128d __B, __m128d __C) {
return _mm_set_sd(__builtin_elementwise_fma(-__A[0], __B[0], __C[0]));
}
static __inline__ __m128 __DEFAULT_FN_ATTRS128_CONSTEXPR
_mm_nmsub_ps(__m128 __A, __m128 __B, __m128 __C) {
return (__m128)__builtin_elementwise_fma(-(__v4sf)__A, (__v4sf)__B,
-(__v4sf)__C);
}
static __inline__ __m128d __DEFAULT_FN_ATTRS128_CONSTEXPR
_mm_nmsub_pd(__m128d __A, __m128d __B, __m128d __C) {
return (__m128d)__builtin_elementwise_fma(-(__v2df)__A, (__v2df)__B,
-(__v2df)__C);
}
static __inline__ __m128 __DEFAULT_FN_ATTRS128_CONSTEXPR
_mm_nmsub_ss(__m128 __A, __m128 __B, __m128 __C) {
return _mm_set_ss(__builtin_elementwise_fma(-__A[0], __B[0], -__C[0]));
}
static __inline__ __m128d __DEFAULT_FN_ATTRS128_CONSTEXPR
_mm_nmsub_sd(__m128d __A, __m128d __B, __m128d __C) {
return _mm_set_sd(__builtin_elementwise_fma(-__A[0], __B[0], -__C[0]));
}
static __inline__ __m128 __DEFAULT_FN_ATTRS128
_mm_maddsub_ps(__m128 __A, __m128 __B, __m128 __C)
{
return (__m128)__builtin_ia32_vfmaddsubps((__v4sf)__A, (__v4sf)__B, (__v4sf)__C);
}
static __inline__ __m128d __DEFAULT_FN_ATTRS128
_mm_maddsub_pd(__m128d __A, __m128d __B, __m128d __C)
{
return (__m128d)__builtin_ia32_vfmaddsubpd((__v2df)__A, (__v2df)__B, (__v2df)__C);
}
static __inline__ __m128 __DEFAULT_FN_ATTRS128
_mm_msubadd_ps(__m128 __A, __m128 __B, __m128 __C)
{
return (__m128)__builtin_ia32_vfmaddsubps((__v4sf)__A, (__v4sf)__B, -(__v4sf)__C);
}
static __inline__ __m128d __DEFAULT_FN_ATTRS128
_mm_msubadd_pd(__m128d __A, __m128d __B, __m128d __C)
{
return (__m128d)__builtin_ia32_vfmaddsubpd((__v2df)__A, (__v2df)__B, -(__v2df)__C);
}
static __inline__ __m256 __DEFAULT_FN_ATTRS256_CONSTEXPR
_mm256_macc_ps(__m256 __A, __m256 __B, __m256 __C) {
return (__m256)__builtin_elementwise_fma((__v8sf)__A, (__v8sf)__B,
(__v8sf)__C);
}
static __inline__ __m256d __DEFAULT_FN_ATTRS256_CONSTEXPR
_mm256_macc_pd(__m256d __A, __m256d __B, __m256d __C) {
return (__m256d)__builtin_elementwise_fma((__v4df)__A, (__v4df)__B,
(__v4df)__C);
}
static __inline__ __m256 __DEFAULT_FN_ATTRS256_CONSTEXPR
_mm256_msub_ps(__m256 __A, __m256 __B, __m256 __C) {
return (__m256)__builtin_elementwise_fma((__v8sf)__A, (__v8sf)__B,
-(__v8sf)__C);
}
static __inline__ __m256d __DEFAULT_FN_ATTRS256_CONSTEXPR
_mm256_msub_pd(__m256d __A, __m256d __B, __m256d __C) {
return (__m256d)__builtin_elementwise_fma((__v4df)__A, (__v4df)__B,
-(__v4df)__C);
}
static __inline__ __m256 __DEFAULT_FN_ATTRS256_CONSTEXPR
_mm256_nmacc_ps(__m256 __A, __m256 __B, __m256 __C) {
return (__m256)__builtin_elementwise_fma(-(__v8sf)__A, (__v8sf)__B,
(__v8sf)__C);
}
static __inline__ __m256d __DEFAULT_FN_ATTRS256_CONSTEXPR
_mm256_nmacc_pd(__m256d __A, __m256d __B, __m256d __C) {
return (__m256d)__builtin_elementwise_fma(-(__v4df)__A, (__v4df)__B,
(__v4df)__C);
}
static __inline__ __m256 __DEFAULT_FN_ATTRS256_CONSTEXPR
_mm256_nmsub_ps(__m256 __A, __m256 __B, __m256 __C) {
return (__m256)__builtin_elementwise_fma(-(__v8sf)__A, (__v8sf)__B,
-(__v8sf)__C);
}
static __inline__ __m256d __DEFAULT_FN_ATTRS256_CONSTEXPR
_mm256_nmsub_pd(__m256d __A, __m256d __B, __m256d __C) {
return (__m256d)__builtin_elementwise_fma(-(__v4df)__A, (__v4df)__B,
-(__v4df)__C);
}
static __inline__ __m256 __DEFAULT_FN_ATTRS256
_mm256_maddsub_ps(__m256 __A, __m256 __B, __m256 __C)
{
return (__m256)__builtin_ia32_vfmaddsubps256((__v8sf)__A, (__v8sf)__B, (__v8sf)__C);
}
static __inline__ __m256d __DEFAULT_FN_ATTRS256
_mm256_maddsub_pd(__m256d __A, __m256d __B, __m256d __C)
{
return (__m256d)__builtin_ia32_vfmaddsubpd256((__v4df)__A, (__v4df)__B, (__v4df)__C);
}
static __inline__ __m256 __DEFAULT_FN_ATTRS256
_mm256_msubadd_ps(__m256 __A, __m256 __B, __m256 __C)
{
return (__m256)__builtin_ia32_vfmaddsubps256((__v8sf)__A, (__v8sf)__B, -(__v8sf)__C);
}
static __inline__ __m256d __DEFAULT_FN_ATTRS256
_mm256_msubadd_pd(__m256d __A, __m256d __B, __m256d __C)
{
return (__m256d)__builtin_ia32_vfmaddsubpd256((__v4df)__A, (__v4df)__B, -(__v4df)__C);
}
#undef __DEFAULT_FN_ATTRS128
#undef __DEFAULT_FN_ATTRS256
#undef __DEFAULT_FN_ATTRS128_CONSTEXPR
#undef __DEFAULT_FN_ATTRS256_CONSTEXPR
#endif /* __FMA4INTRIN_H */
Reference in New Issue View Git Blame Copy Permalink