fma4intrin.h revision 1.12
1 1.12 mrg /* Copyright (C) 2007-2022 Free Software Foundation, Inc. 2 1.1 mrg 3 1.1 mrg This file is part of GCC. 4 1.1 mrg 5 1.1 mrg GCC is free software; you can redistribute it and/or modify 6 1.1 mrg it under the terms of the GNU General Public License as published by 7 1.1 mrg the Free Software Foundation; either version 3, or (at your option) 8 1.1 mrg any later version. 9 1.1 mrg 10 1.1 mrg GCC is distributed in the hope that it will be useful, 11 1.1 mrg but WITHOUT ANY WARRANTY; without even the implied warranty of 12 1.1 mrg MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 13 1.1 mrg GNU General Public License for more details. 14 1.1 mrg 15 1.1 mrg Under Section 7 of GPL version 3, you are granted additional 16 1.1 mrg permissions described in the GCC Runtime Library Exception, version 17 1.1 mrg 3.1, as published by the Free Software Foundation. 18 1.1 mrg 19 1.1 mrg You should have received a copy of the GNU General Public License and 20 1.1 mrg a copy of the GCC Runtime Library Exception along with this program; 21 1.1 mrg see the files COPYING3 and COPYING.RUNTIME respectively. If not, see 22 1.1 mrg <http://www.gnu.org/licenses/>. */ 23 1.1 mrg 24 1.1 mrg #ifndef _X86INTRIN_H_INCLUDED 25 1.1 mrg # error "Never use <fma4intrin.h> directly; include <x86intrin.h> instead." 26 1.1 mrg #endif 27 1.1 mrg 28 1.1 mrg #ifndef _FMA4INTRIN_H_INCLUDED 29 1.1 mrg #define _FMA4INTRIN_H_INCLUDED 30 1.1 mrg 31 1.1 mrg /* We need definitions from the SSE4A, SSE3, SSE2 and SSE header files. */ 32 1.1 mrg #include <ammintrin.h> 33 1.1 mrg 34 1.5 mrg #ifndef __FMA4__ 35 1.5 mrg #pragma GCC push_options 36 1.5 mrg #pragma GCC target("fma4") 37 1.5 mrg #define __DISABLE_FMA4__ 38 1.5 mrg #endif /* __FMA4__ */ 39 1.5 mrg 40 1.1 mrg /* 128b Floating point multiply/add type instructions. */ 41 1.1 mrg extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 42 1.1 mrg _mm_macc_ps (__m128 __A, __m128 __B, __m128 __C) 43 1.1 mrg { 44 1.1 mrg return (__m128) __builtin_ia32_vfmaddps ((__v4sf)__A, (__v4sf)__B, (__v4sf)__C); 45 1.1 mrg } 46 1.1 mrg 47 1.1 mrg extern __inline __m128d __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 48 1.1 mrg _mm_macc_pd (__m128d __A, __m128d __B, __m128d __C) 49 1.1 mrg { 50 1.1 mrg return (__m128d) __builtin_ia32_vfmaddpd ((__v2df)__A, (__v2df)__B, (__v2df)__C); 51 1.1 mrg } 52 1.1 mrg 53 1.1 mrg extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 54 1.1 mrg _mm_macc_ss (__m128 __A, __m128 __B, __m128 __C) 55 1.1 mrg { 56 1.1 mrg return (__m128) __builtin_ia32_vfmaddss ((__v4sf)__A, (__v4sf)__B, (__v4sf)__C); 57 1.1 mrg } 58 1.1 mrg 59 1.1 mrg extern __inline __m128d __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 60 1.1 mrg _mm_macc_sd (__m128d __A, __m128d __B, __m128d __C) 61 1.1 mrg { 62 1.1 mrg return (__m128d) __builtin_ia32_vfmaddsd ((__v2df)__A, (__v2df)__B, (__v2df)__C); 63 1.1 mrg } 64 1.1 mrg 65 1.1 mrg extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 66 1.1 mrg _mm_msub_ps (__m128 __A, __m128 __B, __m128 __C) 67 1.1 mrg 68 1.1 mrg { 69 1.3 mrg return (__m128) __builtin_ia32_vfmaddps ((__v4sf)__A, (__v4sf)__B, -(__v4sf)__C); 70 1.1 mrg } 71 1.1 mrg 72 1.1 mrg extern __inline __m128d __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 73 1.1 mrg _mm_msub_pd (__m128d __A, __m128d __B, __m128d __C) 74 1.1 mrg { 75 1.3 mrg return (__m128d) __builtin_ia32_vfmaddpd ((__v2df)__A, (__v2df)__B, -(__v2df)__C); 76 1.1 mrg } 77 1.1 mrg 78 1.1 mrg extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 79 1.1 mrg _mm_msub_ss (__m128 __A, __m128 __B, __m128 __C) 80 1.1 mrg { 81 1.3 mrg return (__m128) __builtin_ia32_vfmaddss ((__v4sf)__A, (__v4sf)__B, -(__v4sf)__C); 82 1.1 mrg } 83 1.1 mrg 84 1.1 mrg extern __inline __m128d __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 85 1.1 mrg _mm_msub_sd (__m128d __A, __m128d __B, __m128d __C) 86 1.1 mrg { 87 1.3 mrg return (__m128d) __builtin_ia32_vfmaddsd ((__v2df)__A, (__v2df)__B, -(__v2df)__C); 88 1.1 mrg } 89 1.1 mrg 90 1.1 mrg extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 91 1.1 mrg _mm_nmacc_ps (__m128 __A, __m128 __B, __m128 __C) 92 1.1 mrg { 93 1.3 mrg return (__m128) __builtin_ia32_vfmaddps (-(__v4sf)__A, (__v4sf)__B, (__v4sf)__C); 94 1.1 mrg } 95 1.1 mrg 96 1.1 mrg extern __inline __m128d __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 97 1.1 mrg _mm_nmacc_pd (__m128d __A, __m128d __B, __m128d __C) 98 1.1 mrg { 99 1.3 mrg return (__m128d) __builtin_ia32_vfmaddpd (-(__v2df)__A, (__v2df)__B, (__v2df)__C); 100 1.1 mrg } 101 1.1 mrg 102 1.1 mrg extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 103 1.1 mrg _mm_nmacc_ss (__m128 __A, __m128 __B, __m128 __C) 104 1.1 mrg { 105 1.3 mrg return (__m128) __builtin_ia32_vfmaddss (-(__v4sf)__A, (__v4sf)__B, (__v4sf)__C); 106 1.1 mrg } 107 1.1 mrg 108 1.1 mrg extern __inline __m128d __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 109 1.1 mrg _mm_nmacc_sd (__m128d __A, __m128d __B, __m128d __C) 110 1.1 mrg { 111 1.3 mrg return (__m128d) __builtin_ia32_vfmaddsd (-(__v2df)__A, (__v2df)__B, (__v2df)__C); 112 1.1 mrg } 113 1.1 mrg 114 1.1 mrg extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 115 1.1 mrg _mm_nmsub_ps (__m128 __A, __m128 __B, __m128 __C) 116 1.1 mrg { 117 1.3 mrg return (__m128) __builtin_ia32_vfmaddps (-(__v4sf)__A, (__v4sf)__B, -(__v4sf)__C); 118 1.1 mrg } 119 1.1 mrg 120 1.1 mrg extern __inline __m128d __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 121 1.1 mrg _mm_nmsub_pd (__m128d __A, __m128d __B, __m128d __C) 122 1.1 mrg { 123 1.3 mrg return (__m128d) __builtin_ia32_vfmaddpd (-(__v2df)__A, (__v2df)__B, -(__v2df)__C); 124 1.1 mrg } 125 1.1 mrg 126 1.1 mrg extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 127 1.1 mrg _mm_nmsub_ss (__m128 __A, __m128 __B, __m128 __C) 128 1.1 mrg { 129 1.3 mrg return (__m128) __builtin_ia32_vfmaddss (-(__v4sf)__A, (__v4sf)__B, -(__v4sf)__C); 130 1.1 mrg } 131 1.1 mrg 132 1.1 mrg extern __inline __m128d __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 133 1.1 mrg _mm_nmsub_sd (__m128d __A, __m128d __B, __m128d __C) 134 1.1 mrg { 135 1.3 mrg return (__m128d) __builtin_ia32_vfmaddsd (-(__v2df)__A, (__v2df)__B, -(__v2df)__C); 136 1.1 mrg } 137 1.1 mrg 138 1.1 mrg extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 139 1.1 mrg _mm_maddsub_ps (__m128 __A, __m128 __B, __m128 __C) 140 1.1 mrg { 141 1.1 mrg return (__m128) __builtin_ia32_vfmaddsubps ((__v4sf)__A, (__v4sf)__B, (__v4sf)__C); 142 1.1 mrg } 143 1.1 mrg 144 1.1 mrg extern __inline __m128d __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 145 1.1 mrg _mm_maddsub_pd (__m128d __A, __m128d __B, __m128d __C) 146 1.1 mrg { 147 1.1 mrg return (__m128d) __builtin_ia32_vfmaddsubpd ((__v2df)__A, (__v2df)__B, (__v2df)__C); 148 1.1 mrg } 149 1.1 mrg 150 1.1 mrg extern __inline __m128 __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 151 1.1 mrg _mm_msubadd_ps (__m128 __A, __m128 __B, __m128 __C) 152 1.1 mrg { 153 1.3 mrg return (__m128) __builtin_ia32_vfmaddsubps ((__v4sf)__A, (__v4sf)__B, -(__v4sf)__C); 154 1.1 mrg } 155 1.1 mrg 156 1.1 mrg extern __inline __m128d __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 157 1.1 mrg _mm_msubadd_pd (__m128d __A, __m128d __B, __m128d __C) 158 1.1 mrg { 159 1.3 mrg return (__m128d) __builtin_ia32_vfmaddsubpd ((__v2df)__A, (__v2df)__B, -(__v2df)__C); 160 1.1 mrg } 161 1.1 mrg 162 1.1 mrg /* 256b Floating point multiply/add type instructions. */ 163 1.1 mrg extern __inline __m256 __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 164 1.1 mrg _mm256_macc_ps (__m256 __A, __m256 __B, __m256 __C) 165 1.1 mrg { 166 1.1 mrg return (__m256) __builtin_ia32_vfmaddps256 ((__v8sf)__A, (__v8sf)__B, (__v8sf)__C); 167 1.1 mrg } 168 1.1 mrg 169 1.1 mrg extern __inline __m256d __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 170 1.1 mrg _mm256_macc_pd (__m256d __A, __m256d __B, __m256d __C) 171 1.1 mrg { 172 1.1 mrg return (__m256d) __builtin_ia32_vfmaddpd256 ((__v4df)__A, (__v4df)__B, (__v4df)__C); 173 1.1 mrg } 174 1.1 mrg 175 1.1 mrg extern __inline __m256 __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 176 1.1 mrg _mm256_msub_ps (__m256 __A, __m256 __B, __m256 __C) 177 1.1 mrg 178 1.1 mrg { 179 1.3 mrg return (__m256) __builtin_ia32_vfmaddps256 ((__v8sf)__A, (__v8sf)__B, -(__v8sf)__C); 180 1.1 mrg } 181 1.1 mrg 182 1.1 mrg extern __inline __m256d __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 183 1.1 mrg _mm256_msub_pd (__m256d __A, __m256d __B, __m256d __C) 184 1.1 mrg { 185 1.3 mrg return (__m256d) __builtin_ia32_vfmaddpd256 ((__v4df)__A, (__v4df)__B, -(__v4df)__C); 186 1.1 mrg } 187 1.1 mrg 188 1.1 mrg extern __inline __m256 __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 189 1.1 mrg _mm256_nmacc_ps (__m256 __A, __m256 __B, __m256 __C) 190 1.1 mrg { 191 1.3 mrg return (__m256) __builtin_ia32_vfmaddps256 (-(__v8sf)__A, (__v8sf)__B, (__v8sf)__C); 192 1.1 mrg } 193 1.1 mrg 194 1.1 mrg extern __inline __m256d __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 195 1.1 mrg _mm256_nmacc_pd (__m256d __A, __m256d __B, __m256d __C) 196 1.1 mrg { 197 1.3 mrg return (__m256d) __builtin_ia32_vfmaddpd256 (-(__v4df)__A, (__v4df)__B, (__v4df)__C); 198 1.1 mrg } 199 1.1 mrg 200 1.1 mrg extern __inline __m256 __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 201 1.1 mrg _mm256_nmsub_ps (__m256 __A, __m256 __B, __m256 __C) 202 1.1 mrg { 203 1.3 mrg return (__m256) __builtin_ia32_vfmaddps256 (-(__v8sf)__A, (__v8sf)__B, -(__v8sf)__C); 204 1.1 mrg } 205 1.1 mrg 206 1.1 mrg extern __inline __m256d __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 207 1.1 mrg _mm256_nmsub_pd (__m256d __A, __m256d __B, __m256d __C) 208 1.1 mrg { 209 1.3 mrg return (__m256d) __builtin_ia32_vfmaddpd256 (-(__v4df)__A, (__v4df)__B, -(__v4df)__C); 210 1.1 mrg } 211 1.1 mrg 212 1.1 mrg extern __inline __m256 __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 213 1.1 mrg _mm256_maddsub_ps (__m256 __A, __m256 __B, __m256 __C) 214 1.1 mrg { 215 1.1 mrg return (__m256) __builtin_ia32_vfmaddsubps256 ((__v8sf)__A, (__v8sf)__B, (__v8sf)__C); 216 1.1 mrg } 217 1.1 mrg 218 1.1 mrg extern __inline __m256d __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 219 1.1 mrg _mm256_maddsub_pd (__m256d __A, __m256d __B, __m256d __C) 220 1.1 mrg { 221 1.1 mrg return (__m256d) __builtin_ia32_vfmaddsubpd256 ((__v4df)__A, (__v4df)__B, (__v4df)__C); 222 1.1 mrg } 223 1.1 mrg 224 1.1 mrg extern __inline __m256 __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 225 1.1 mrg _mm256_msubadd_ps (__m256 __A, __m256 __B, __m256 __C) 226 1.1 mrg { 227 1.3 mrg return (__m256) __builtin_ia32_vfmaddsubps256 ((__v8sf)__A, (__v8sf)__B, -(__v8sf)__C); 228 1.1 mrg } 229 1.1 mrg 230 1.1 mrg extern __inline __m256d __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 231 1.1 mrg _mm256_msubadd_pd (__m256d __A, __m256d __B, __m256d __C) 232 1.1 mrg { 233 1.3 mrg return (__m256d) __builtin_ia32_vfmaddsubpd256 ((__v4df)__A, (__v4df)__B, -(__v4df)__C); 234 1.1 mrg } 235 1.1 mrg 236 1.5 mrg #ifdef __DISABLE_FMA4__ 237 1.5 mrg #undef __DISABLE_FMA4__ 238 1.5 mrg #pragma GCC pop_options 239 1.5 mrg #endif /* __DISABLE_FMA4__ */ 240 1.1 mrg 241 1.1 mrg #endif 242
Indexes created Mon Apr 20 00:23:12 UTC 2026