Ying Wang | 6099914 | 2013-01-07 13:59:36 -0800 | [diff] [blame^] | 1 | /*===---- fma4intrin.h - FMA4 intrinsics -----------------------------------=== |
| 2 | * |
| 3 | * Permission is hereby granted, free of charge, to any person obtaining a copy |
| 4 | * of this software and associated documentation files (the "Software"), to deal |
| 5 | * in the Software without restriction, including without limitation the rights |
| 6 | * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell |
| 7 | * copies of the Software, and to permit persons to whom the Software is |
| 8 | * furnished to do so, subject to the following conditions: |
| 9 | * |
| 10 | * The above copyright notice and this permission notice shall be included in |
| 11 | * all copies or substantial portions of the Software. |
| 12 | * |
| 13 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
| 14 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
| 15 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE |
| 16 | * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
| 17 | * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
| 18 | * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN |
| 19 | * THE SOFTWARE. |
| 20 | * |
| 21 | *===-----------------------------------------------------------------------=== |
| 22 | */ |
| 23 | |
| 24 | #ifndef __X86INTRIN_H |
| 25 | #error "Never use <fma4intrin.h> directly; include <x86intrin.h> instead." |
| 26 | #endif |
| 27 | |
| 28 | #ifndef __FMA4INTRIN_H |
| 29 | #define __FMA4INTRIN_H |
| 30 | |
| 31 | #ifndef __FMA4__ |
| 32 | # error "FMA4 instruction set is not enabled" |
| 33 | #else |
| 34 | |
| 35 | #include <pmmintrin.h> |
| 36 | |
| 37 | static __inline__ __m128 __attribute__((__always_inline__, __nodebug__)) |
| 38 | _mm_macc_ps(__m128 __A, __m128 __B, __m128 __C) |
| 39 | { |
| 40 | return (__m128)__builtin_ia32_vfmaddps(__A, __B, __C); |
| 41 | } |
| 42 | |
| 43 | static __inline__ __m128d __attribute__((__always_inline__, __nodebug__)) |
| 44 | _mm_macc_pd(__m128d __A, __m128d __B, __m128d __C) |
| 45 | { |
| 46 | return (__m128d)__builtin_ia32_vfmaddpd(__A, __B, __C); |
| 47 | } |
| 48 | |
| 49 | static __inline__ __m128 __attribute__((__always_inline__, __nodebug__)) |
| 50 | _mm_macc_ss(__m128 __A, __m128 __B, __m128 __C) |
| 51 | { |
| 52 | return (__m128)__builtin_ia32_vfmaddss(__A, __B, __C); |
| 53 | } |
| 54 | |
| 55 | static __inline__ __m128d __attribute__((__always_inline__, __nodebug__)) |
| 56 | _mm_macc_sd(__m128d __A, __m128d __B, __m128d __C) |
| 57 | { |
| 58 | return (__m128d)__builtin_ia32_vfmaddsd(__A, __B, __C); |
| 59 | } |
| 60 | |
| 61 | static __inline__ __m128 __attribute__((__always_inline__, __nodebug__)) |
| 62 | _mm_msub_ps(__m128 __A, __m128 __B, __m128 __C) |
| 63 | { |
| 64 | return (__m128)__builtin_ia32_vfmsubps(__A, __B, __C); |
| 65 | } |
| 66 | |
| 67 | static __inline__ __m128d __attribute__((__always_inline__, __nodebug__)) |
| 68 | _mm_msub_pd(__m128d __A, __m128d __B, __m128d __C) |
| 69 | { |
| 70 | return (__m128d)__builtin_ia32_vfmsubpd(__A, __B, __C); |
| 71 | } |
| 72 | |
| 73 | static __inline__ __m128 __attribute__((__always_inline__, __nodebug__)) |
| 74 | _mm_msub_ss(__m128 __A, __m128 __B, __m128 __C) |
| 75 | { |
| 76 | return (__m128)__builtin_ia32_vfmsubss(__A, __B, __C); |
| 77 | } |
| 78 | |
| 79 | static __inline__ __m128d __attribute__((__always_inline__, __nodebug__)) |
| 80 | _mm_msub_sd(__m128d __A, __m128d __B, __m128d __C) |
| 81 | { |
| 82 | return (__m128d)__builtin_ia32_vfmsubsd(__A, __B, __C); |
| 83 | } |
| 84 | |
| 85 | static __inline__ __m128 __attribute__((__always_inline__, __nodebug__)) |
| 86 | _mm_nmacc_ps(__m128 __A, __m128 __B, __m128 __C) |
| 87 | { |
| 88 | return (__m128)__builtin_ia32_vfnmaddps(__A, __B, __C); |
| 89 | } |
| 90 | |
| 91 | static __inline__ __m128d __attribute__((__always_inline__, __nodebug__)) |
| 92 | _mm_nmacc_pd(__m128d __A, __m128d __B, __m128d __C) |
| 93 | { |
| 94 | return (__m128d)__builtin_ia32_vfnmaddpd(__A, __B, __C); |
| 95 | } |
| 96 | |
| 97 | static __inline__ __m128 __attribute__((__always_inline__, __nodebug__)) |
| 98 | _mm_nmacc_ss(__m128 __A, __m128 __B, __m128 __C) |
| 99 | { |
| 100 | return (__m128)__builtin_ia32_vfnmaddss(__A, __B, __C); |
| 101 | } |
| 102 | |
| 103 | static __inline__ __m128d __attribute__((__always_inline__, __nodebug__)) |
| 104 | _mm_nmacc_sd(__m128d __A, __m128d __B, __m128d __C) |
| 105 | { |
| 106 | return (__m128d)__builtin_ia32_vfnmaddsd(__A, __B, __C); |
| 107 | } |
| 108 | |
| 109 | static __inline__ __m128 __attribute__((__always_inline__, __nodebug__)) |
| 110 | _mm_nmsub_ps(__m128 __A, __m128 __B, __m128 __C) |
| 111 | { |
| 112 | return (__m128)__builtin_ia32_vfnmsubps(__A, __B, __C); |
| 113 | } |
| 114 | |
| 115 | static __inline__ __m128d __attribute__((__always_inline__, __nodebug__)) |
| 116 | _mm_nmsub_pd(__m128d __A, __m128d __B, __m128d __C) |
| 117 | { |
| 118 | return (__m128d)__builtin_ia32_vfnmsubpd(__A, __B, __C); |
| 119 | } |
| 120 | |
| 121 | static __inline__ __m128 __attribute__((__always_inline__, __nodebug__)) |
| 122 | _mm_nmsub_ss(__m128 __A, __m128 __B, __m128 __C) |
| 123 | { |
| 124 | return (__m128)__builtin_ia32_vfnmsubss(__A, __B, __C); |
| 125 | } |
| 126 | |
| 127 | static __inline__ __m128d __attribute__((__always_inline__, __nodebug__)) |
| 128 | _mm_nmsub_sd(__m128d __A, __m128d __B, __m128d __C) |
| 129 | { |
| 130 | return (__m128d)__builtin_ia32_vfnmsubsd(__A, __B, __C); |
| 131 | } |
| 132 | |
| 133 | static __inline__ __m128 __attribute__((__always_inline__, __nodebug__)) |
| 134 | _mm_maddsub_ps(__m128 __A, __m128 __B, __m128 __C) |
| 135 | { |
| 136 | return (__m128)__builtin_ia32_vfmaddsubps(__A, __B, __C); |
| 137 | } |
| 138 | |
| 139 | static __inline__ __m128d __attribute__((__always_inline__, __nodebug__)) |
| 140 | _mm_maddsub_pd(__m128d __A, __m128d __B, __m128d __C) |
| 141 | { |
| 142 | return (__m128d)__builtin_ia32_vfmaddsubpd(__A, __B, __C); |
| 143 | } |
| 144 | |
| 145 | static __inline__ __m128 __attribute__((__always_inline__, __nodebug__)) |
| 146 | _mm_msubadd_ps(__m128 __A, __m128 __B, __m128 __C) |
| 147 | { |
| 148 | return (__m128)__builtin_ia32_vfmsubaddps(__A, __B, __C); |
| 149 | } |
| 150 | |
| 151 | static __inline__ __m128d __attribute__((__always_inline__, __nodebug__)) |
| 152 | _mm_msubadd_pd(__m128d __A, __m128d __B, __m128d __C) |
| 153 | { |
| 154 | return (__m128d)__builtin_ia32_vfmsubaddpd(__A, __B, __C); |
| 155 | } |
| 156 | |
| 157 | static __inline__ __m256 __attribute__((__always_inline__, __nodebug__)) |
| 158 | _mm256_macc_ps(__m256 __A, __m256 __B, __m256 __C) |
| 159 | { |
| 160 | return (__m256)__builtin_ia32_vfmaddps256(__A, __B, __C); |
| 161 | } |
| 162 | |
| 163 | static __inline__ __m256d __attribute__((__always_inline__, __nodebug__)) |
| 164 | _mm256_macc_pd(__m256d __A, __m256d __B, __m256d __C) |
| 165 | { |
| 166 | return (__m256d)__builtin_ia32_vfmaddpd256(__A, __B, __C); |
| 167 | } |
| 168 | |
| 169 | static __inline__ __m256 __attribute__((__always_inline__, __nodebug__)) |
| 170 | _mm256_msub_ps(__m256 __A, __m256 __B, __m256 __C) |
| 171 | { |
| 172 | return (__m256)__builtin_ia32_vfmsubps256(__A, __B, __C); |
| 173 | } |
| 174 | |
| 175 | static __inline__ __m256d __attribute__((__always_inline__, __nodebug__)) |
| 176 | _mm256_msub_pd(__m256d __A, __m256d __B, __m256d __C) |
| 177 | { |
| 178 | return (__m256d)__builtin_ia32_vfmsubpd256(__A, __B, __C); |
| 179 | } |
| 180 | |
| 181 | static __inline__ __m256 __attribute__((__always_inline__, __nodebug__)) |
| 182 | _mm256_nmacc_ps(__m256 __A, __m256 __B, __m256 __C) |
| 183 | { |
| 184 | return (__m256)__builtin_ia32_vfnmaddps256(__A, __B, __C); |
| 185 | } |
| 186 | |
| 187 | static __inline__ __m256d __attribute__((__always_inline__, __nodebug__)) |
| 188 | _mm256_nmacc_pd(__m256d __A, __m256d __B, __m256d __C) |
| 189 | { |
| 190 | return (__m256d)__builtin_ia32_vfnmaddpd256(__A, __B, __C); |
| 191 | } |
| 192 | |
| 193 | static __inline__ __m256 __attribute__((__always_inline__, __nodebug__)) |
| 194 | _mm256_nmsub_ps(__m256 __A, __m256 __B, __m256 __C) |
| 195 | { |
| 196 | return (__m256)__builtin_ia32_vfnmsubps256(__A, __B, __C); |
| 197 | } |
| 198 | |
| 199 | static __inline__ __m256d __attribute__((__always_inline__, __nodebug__)) |
| 200 | _mm256_nmsub_pd(__m256d __A, __m256d __B, __m256d __C) |
| 201 | { |
| 202 | return (__m256d)__builtin_ia32_vfnmsubpd256(__A, __B, __C); |
| 203 | } |
| 204 | |
| 205 | static __inline__ __m256 __attribute__((__always_inline__, __nodebug__)) |
| 206 | _mm256_maddsub_ps(__m256 __A, __m256 __B, __m256 __C) |
| 207 | { |
| 208 | return (__m256)__builtin_ia32_vfmaddsubps256(__A, __B, __C); |
| 209 | } |
| 210 | |
| 211 | static __inline__ __m256d __attribute__((__always_inline__, __nodebug__)) |
| 212 | _mm256_maddsub_pd(__m256d __A, __m256d __B, __m256d __C) |
| 213 | { |
| 214 | return (__m256d)__builtin_ia32_vfmaddsubpd256(__A, __B, __C); |
| 215 | } |
| 216 | |
| 217 | static __inline__ __m256 __attribute__((__always_inline__, __nodebug__)) |
| 218 | _mm256_msubadd_ps(__m256 __A, __m256 __B, __m256 __C) |
| 219 | { |
| 220 | return (__m256)__builtin_ia32_vfmsubaddps256(__A, __B, __C); |
| 221 | } |
| 222 | |
| 223 | static __inline__ __m256d __attribute__((__always_inline__, __nodebug__)) |
| 224 | _mm256_msubadd_pd(__m256d __A, __m256d __B, __m256d __C) |
| 225 | { |
| 226 | return (__m256d)__builtin_ia32_vfmsubaddpd256(__A, __B, __C); |
| 227 | } |
| 228 | |
| 229 | #endif /* __FMA4__ */ |
| 230 | |
| 231 | #endif /* __FMA4INTRIN_H */ |