| 1 |
709 |
jeremybenn |
/* Copyright (C) 2007, 2008, 2009, 2010 Free Software Foundation, Inc.
|
| 2 |
|
|
|
| 3 |
|
|
This file is part of GCC.
|
| 4 |
|
|
|
| 5 |
|
|
GCC is free software; you can redistribute it and/or modify
|
| 6 |
|
|
it under the terms of the GNU General Public License as published by
|
| 7 |
|
|
the Free Software Foundation; either version 3, or (at your option)
|
| 8 |
|
|
any later version.
|
| 9 |
|
|
|
| 10 |
|
|
GCC is distributed in the hope that it will be useful,
|
| 11 |
|
|
but WITHOUT ANY WARRANTY; without even the implied warranty of
|
| 12 |
|
|
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
| 13 |
|
|
GNU General Public License for more details.
|
| 14 |
|
|
|
| 15 |
|
|
Under Section 7 of GPL version 3, you are granted additional
|
| 16 |
|
|
permissions described in the GCC Runtime Library Exception, version
|
| 17 |
|
|
3.1, as published by the Free Software Foundation.
|
| 18 |
|
|
|
| 19 |
|
|
You should have received a copy of the GNU General Public License and
|
| 20 |
|
|
a copy of the GCC Runtime Library Exception along with this program;
|
| 21 |
|
|
see the files COPYING3 and COPYING.RUNTIME respectively. If not, see
|
| 22 |
|
|
<http://www.gnu.org/licenses/>. */
|
| 23 |
|
|
|
| 24 |
|
|
#ifndef _X86INTRIN_H_INCLUDED
|
| 25 |
|
|
# error "Never use <fma4intrin.h> directly; include <x86intrin.h> instead."
|
| 26 |
|
|
#endif
|
| 27 |
|
|
|
| 28 |
|
|
#ifndef _FMA4INTRIN_H_INCLUDED
|
| 29 |
|
|
#define _FMA4INTRIN_H_INCLUDED
|
| 30 |
|
|
|
| 31 |
|
|
#ifndef __FMA4__
|
| 32 |
|
|
# error "FMA4 instruction set not enabled"
|
| 33 |
|
|
#else
|
| 34 |
|
|
|
| 35 |
|
|
/* We need definitions from the SSE4A, SSE3, SSE2 and SSE header files. */
|
| 36 |
|
|
#include <ammintrin.h>
|
| 37 |
|
|
|
| 38 |
|
|
/* 128b Floating point multiply/add type instructions. */
|
| 39 |
|
|
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
| 40 |
|
|
_mm_macc_ps (__m128 __A, __m128 __B, __m128 __C)
|
| 41 |
|
|
{
|
| 42 |
|
|
return (__m128) __builtin_ia32_vfmaddps ((__v4sf)__A, (__v4sf)__B, (__v4sf)__C);
|
| 43 |
|
|
}
|
| 44 |
|
|
|
| 45 |
|
|
extern __inline __m128d __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
| 46 |
|
|
_mm_macc_pd (__m128d __A, __m128d __B, __m128d __C)
|
| 47 |
|
|
{
|
| 48 |
|
|
return (__m128d) __builtin_ia32_vfmaddpd ((__v2df)__A, (__v2df)__B, (__v2df)__C);
|
| 49 |
|
|
}
|
| 50 |
|
|
|
| 51 |
|
|
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
| 52 |
|
|
_mm_macc_ss (__m128 __A, __m128 __B, __m128 __C)
|
| 53 |
|
|
{
|
| 54 |
|
|
return (__m128) __builtin_ia32_vfmaddss ((__v4sf)__A, (__v4sf)__B, (__v4sf)__C);
|
| 55 |
|
|
}
|
| 56 |
|
|
|
| 57 |
|
|
extern __inline __m128d __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
| 58 |
|
|
_mm_macc_sd (__m128d __A, __m128d __B, __m128d __C)
|
| 59 |
|
|
{
|
| 60 |
|
|
return (__m128d) __builtin_ia32_vfmaddsd ((__v2df)__A, (__v2df)__B, (__v2df)__C);
|
| 61 |
|
|
}
|
| 62 |
|
|
|
| 63 |
|
|
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
| 64 |
|
|
_mm_msub_ps (__m128 __A, __m128 __B, __m128 __C)
|
| 65 |
|
|
|
| 66 |
|
|
{
|
| 67 |
|
|
return (__m128) __builtin_ia32_vfmaddps ((__v4sf)__A, (__v4sf)__B, -(__v4sf)__C);
|
| 68 |
|
|
}
|
| 69 |
|
|
|
| 70 |
|
|
extern __inline __m128d __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
| 71 |
|
|
_mm_msub_pd (__m128d __A, __m128d __B, __m128d __C)
|
| 72 |
|
|
{
|
| 73 |
|
|
return (__m128d) __builtin_ia32_vfmaddpd ((__v2df)__A, (__v2df)__B, -(__v2df)__C);
|
| 74 |
|
|
}
|
| 75 |
|
|
|
| 76 |
|
|
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
| 77 |
|
|
_mm_msub_ss (__m128 __A, __m128 __B, __m128 __C)
|
| 78 |
|
|
{
|
| 79 |
|
|
return (__m128) __builtin_ia32_vfmaddss ((__v4sf)__A, (__v4sf)__B, -(__v4sf)__C);
|
| 80 |
|
|
}
|
| 81 |
|
|
|
| 82 |
|
|
extern __inline __m128d __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
| 83 |
|
|
_mm_msub_sd (__m128d __A, __m128d __B, __m128d __C)
|
| 84 |
|
|
{
|
| 85 |
|
|
return (__m128d) __builtin_ia32_vfmaddsd ((__v2df)__A, (__v2df)__B, -(__v2df)__C);
|
| 86 |
|
|
}
|
| 87 |
|
|
|
| 88 |
|
|
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
| 89 |
|
|
_mm_nmacc_ps (__m128 __A, __m128 __B, __m128 __C)
|
| 90 |
|
|
{
|
| 91 |
|
|
return (__m128) __builtin_ia32_vfmaddps (-(__v4sf)__A, (__v4sf)__B, (__v4sf)__C);
|
| 92 |
|
|
}
|
| 93 |
|
|
|
| 94 |
|
|
extern __inline __m128d __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
| 95 |
|
|
_mm_nmacc_pd (__m128d __A, __m128d __B, __m128d __C)
|
| 96 |
|
|
{
|
| 97 |
|
|
return (__m128d) __builtin_ia32_vfmaddpd (-(__v2df)__A, (__v2df)__B, (__v2df)__C);
|
| 98 |
|
|
}
|
| 99 |
|
|
|
| 100 |
|
|
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
| 101 |
|
|
_mm_nmacc_ss (__m128 __A, __m128 __B, __m128 __C)
|
| 102 |
|
|
{
|
| 103 |
|
|
return (__m128) __builtin_ia32_vfmaddss (-(__v4sf)__A, (__v4sf)__B, (__v4sf)__C);
|
| 104 |
|
|
}
|
| 105 |
|
|
|
| 106 |
|
|
extern __inline __m128d __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
| 107 |
|
|
_mm_nmacc_sd (__m128d __A, __m128d __B, __m128d __C)
|
| 108 |
|
|
{
|
| 109 |
|
|
return (__m128d) __builtin_ia32_vfmaddsd (-(__v2df)__A, (__v2df)__B, (__v2df)__C);
|
| 110 |
|
|
}
|
| 111 |
|
|
|
| 112 |
|
|
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
| 113 |
|
|
_mm_nmsub_ps (__m128 __A, __m128 __B, __m128 __C)
|
| 114 |
|
|
{
|
| 115 |
|
|
return (__m128) __builtin_ia32_vfmaddps (-(__v4sf)__A, (__v4sf)__B, -(__v4sf)__C);
|
| 116 |
|
|
}
|
| 117 |
|
|
|
| 118 |
|
|
extern __inline __m128d __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
| 119 |
|
|
_mm_nmsub_pd (__m128d __A, __m128d __B, __m128d __C)
|
| 120 |
|
|
{
|
| 121 |
|
|
return (__m128d) __builtin_ia32_vfmaddpd (-(__v2df)__A, (__v2df)__B, -(__v2df)__C);
|
| 122 |
|
|
}
|
| 123 |
|
|
|
| 124 |
|
|
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
| 125 |
|
|
_mm_nmsub_ss (__m128 __A, __m128 __B, __m128 __C)
|
| 126 |
|
|
{
|
| 127 |
|
|
return (__m128) __builtin_ia32_vfmaddss (-(__v4sf)__A, (__v4sf)__B, -(__v4sf)__C);
|
| 128 |
|
|
}
|
| 129 |
|
|
|
| 130 |
|
|
extern __inline __m128d __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
| 131 |
|
|
_mm_nmsub_sd (__m128d __A, __m128d __B, __m128d __C)
|
| 132 |
|
|
{
|
| 133 |
|
|
return (__m128d) __builtin_ia32_vfmaddsd (-(__v2df)__A, (__v2df)__B, -(__v2df)__C);
|
| 134 |
|
|
}
|
| 135 |
|
|
|
| 136 |
|
|
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
| 137 |
|
|
_mm_maddsub_ps (__m128 __A, __m128 __B, __m128 __C)
|
| 138 |
|
|
{
|
| 139 |
|
|
return (__m128) __builtin_ia32_vfmaddsubps ((__v4sf)__A, (__v4sf)__B, (__v4sf)__C);
|
| 140 |
|
|
}
|
| 141 |
|
|
|
| 142 |
|
|
extern __inline __m128d __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
| 143 |
|
|
_mm_maddsub_pd (__m128d __A, __m128d __B, __m128d __C)
|
| 144 |
|
|
{
|
| 145 |
|
|
return (__m128d) __builtin_ia32_vfmaddsubpd ((__v2df)__A, (__v2df)__B, (__v2df)__C);
|
| 146 |
|
|
}
|
| 147 |
|
|
|
| 148 |
|
|
extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
| 149 |
|
|
_mm_msubadd_ps (__m128 __A, __m128 __B, __m128 __C)
|
| 150 |
|
|
{
|
| 151 |
|
|
return (__m128) __builtin_ia32_vfmaddsubps ((__v4sf)__A, (__v4sf)__B, -(__v4sf)__C);
|
| 152 |
|
|
}
|
| 153 |
|
|
|
| 154 |
|
|
extern __inline __m128d __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
| 155 |
|
|
_mm_msubadd_pd (__m128d __A, __m128d __B, __m128d __C)
|
| 156 |
|
|
{
|
| 157 |
|
|
return (__m128d) __builtin_ia32_vfmaddsubpd ((__v2df)__A, (__v2df)__B, -(__v2df)__C);
|
| 158 |
|
|
}
|
| 159 |
|
|
|
| 160 |
|
|
/* 256b Floating point multiply/add type instructions. */
|
| 161 |
|
|
extern __inline __m256 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
| 162 |
|
|
_mm256_macc_ps (__m256 __A, __m256 __B, __m256 __C)
|
| 163 |
|
|
{
|
| 164 |
|
|
return (__m256) __builtin_ia32_vfmaddps256 ((__v8sf)__A, (__v8sf)__B, (__v8sf)__C);
|
| 165 |
|
|
}
|
| 166 |
|
|
|
| 167 |
|
|
extern __inline __m256d __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
| 168 |
|
|
_mm256_macc_pd (__m256d __A, __m256d __B, __m256d __C)
|
| 169 |
|
|
{
|
| 170 |
|
|
return (__m256d) __builtin_ia32_vfmaddpd256 ((__v4df)__A, (__v4df)__B, (__v4df)__C);
|
| 171 |
|
|
}
|
| 172 |
|
|
|
| 173 |
|
|
extern __inline __m256 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
| 174 |
|
|
_mm256_msub_ps (__m256 __A, __m256 __B, __m256 __C)
|
| 175 |
|
|
|
| 176 |
|
|
{
|
| 177 |
|
|
return (__m256) __builtin_ia32_vfmaddps256 ((__v8sf)__A, (__v8sf)__B, -(__v8sf)__C);
|
| 178 |
|
|
}
|
| 179 |
|
|
|
| 180 |
|
|
extern __inline __m256d __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
| 181 |
|
|
_mm256_msub_pd (__m256d __A, __m256d __B, __m256d __C)
|
| 182 |
|
|
{
|
| 183 |
|
|
return (__m256d) __builtin_ia32_vfmaddpd256 ((__v4df)__A, (__v4df)__B, -(__v4df)__C);
|
| 184 |
|
|
}
|
| 185 |
|
|
|
| 186 |
|
|
extern __inline __m256 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
| 187 |
|
|
_mm256_nmacc_ps (__m256 __A, __m256 __B, __m256 __C)
|
| 188 |
|
|
{
|
| 189 |
|
|
return (__m256) __builtin_ia32_vfmaddps256 (-(__v8sf)__A, (__v8sf)__B, (__v8sf)__C);
|
| 190 |
|
|
}
|
| 191 |
|
|
|
| 192 |
|
|
extern __inline __m256d __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
| 193 |
|
|
_mm256_nmacc_pd (__m256d __A, __m256d __B, __m256d __C)
|
| 194 |
|
|
{
|
| 195 |
|
|
return (__m256d) __builtin_ia32_vfmaddpd256 (-(__v4df)__A, (__v4df)__B, (__v4df)__C);
|
| 196 |
|
|
}
|
| 197 |
|
|
|
| 198 |
|
|
extern __inline __m256 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
| 199 |
|
|
_mm256_nmsub_ps (__m256 __A, __m256 __B, __m256 __C)
|
| 200 |
|
|
{
|
| 201 |
|
|
return (__m256) __builtin_ia32_vfmaddps256 (-(__v8sf)__A, (__v8sf)__B, -(__v8sf)__C);
|
| 202 |
|
|
}
|
| 203 |
|
|
|
| 204 |
|
|
extern __inline __m256d __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
| 205 |
|
|
_mm256_nmsub_pd (__m256d __A, __m256d __B, __m256d __C)
|
| 206 |
|
|
{
|
| 207 |
|
|
return (__m256d) __builtin_ia32_vfmaddpd256 (-(__v4df)__A, (__v4df)__B, -(__v4df)__C);
|
| 208 |
|
|
}
|
| 209 |
|
|
|
| 210 |
|
|
extern __inline __m256 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
| 211 |
|
|
_mm256_maddsub_ps (__m256 __A, __m256 __B, __m256 __C)
|
| 212 |
|
|
{
|
| 213 |
|
|
return (__m256) __builtin_ia32_vfmaddsubps256 ((__v8sf)__A, (__v8sf)__B, (__v8sf)__C);
|
| 214 |
|
|
}
|
| 215 |
|
|
|
| 216 |
|
|
extern __inline __m256d __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
| 217 |
|
|
_mm256_maddsub_pd (__m256d __A, __m256d __B, __m256d __C)
|
| 218 |
|
|
{
|
| 219 |
|
|
return (__m256d) __builtin_ia32_vfmaddsubpd256 ((__v4df)__A, (__v4df)__B, (__v4df)__C);
|
| 220 |
|
|
}
|
| 221 |
|
|
|
| 222 |
|
|
extern __inline __m256 __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
| 223 |
|
|
_mm256_msubadd_ps (__m256 __A, __m256 __B, __m256 __C)
|
| 224 |
|
|
{
|
| 225 |
|
|
return (__m256) __builtin_ia32_vfmaddsubps256 ((__v8sf)__A, (__v8sf)__B, -(__v8sf)__C);
|
| 226 |
|
|
}
|
| 227 |
|
|
|
| 228 |
|
|
extern __inline __m256d __attribute__((__gnu_inline__, __always_inline__, __artificial__))
|
| 229 |
|
|
_mm256_msubadd_pd (__m256d __A, __m256d __B, __m256d __C)
|
| 230 |
|
|
{
|
| 231 |
|
|
return (__m256d) __builtin_ia32_vfmaddsubpd256 ((__v4df)__A, (__v4df)__B, -(__v4df)__C);
|
| 232 |
|
|
}
|
| 233 |
|
|
|
| 234 |
|
|
#endif
|
| 235 |
|
|
|
| 236 |
|
|
#endif
|
