root / lab4 / .minix-src / include / clang-3.6 / avx2intrin.h @ 13
History | View | Annotate | Download (42.2 KB)
| 1 | 13 | up20180614 | /*===---- avx2intrin.h - AVX2 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 __IMMINTRIN_H
|
||
| 25 | #error "Never use <avx2intrin.h> directly; include <immintrin.h> instead." |
||
| 26 | #endif
|
||
| 27 | |||
| 28 | #ifndef __AVX2INTRIN_H
|
||
| 29 | #define __AVX2INTRIN_H
|
||
| 30 | |||
| 31 | /* SSE4 Multiple Packed Sums of Absolute Difference. */
|
||
| 32 | #define _mm256_mpsadbw_epu8(X, Y, M) __builtin_ia32_mpsadbw256((X), (Y), (M))
|
||
| 33 | |||
| 34 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 35 | _mm256_abs_epi8(__m256i __a) |
||
| 36 | {
|
||
| 37 | return (__m256i)__builtin_ia32_pabsb256((__v32qi)__a);
|
||
| 38 | } |
||
| 39 | |||
| 40 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 41 | _mm256_abs_epi16(__m256i __a) |
||
| 42 | {
|
||
| 43 | return (__m256i)__builtin_ia32_pabsw256((__v16hi)__a);
|
||
| 44 | } |
||
| 45 | |||
| 46 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 47 | _mm256_abs_epi32(__m256i __a) |
||
| 48 | {
|
||
| 49 | return (__m256i)__builtin_ia32_pabsd256((__v8si)__a);
|
||
| 50 | } |
||
| 51 | |||
| 52 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 53 | _mm256_packs_epi16(__m256i __a, __m256i __b) |
||
| 54 | {
|
||
| 55 | return (__m256i)__builtin_ia32_packsswb256((__v16hi)__a, (__v16hi)__b);
|
||
| 56 | } |
||
| 57 | |||
| 58 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 59 | _mm256_packs_epi32(__m256i __a, __m256i __b) |
||
| 60 | {
|
||
| 61 | return (__m256i)__builtin_ia32_packssdw256((__v8si)__a, (__v8si)__b);
|
||
| 62 | } |
||
| 63 | |||
| 64 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 65 | _mm256_packus_epi16(__m256i __a, __m256i __b) |
||
| 66 | {
|
||
| 67 | return (__m256i)__builtin_ia32_packuswb256((__v16hi)__a, (__v16hi)__b);
|
||
| 68 | } |
||
| 69 | |||
| 70 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 71 | _mm256_packus_epi32(__m256i __V1, __m256i __V2) |
||
| 72 | {
|
||
| 73 | return (__m256i) __builtin_ia32_packusdw256((__v8si)__V1, (__v8si)__V2);
|
||
| 74 | } |
||
| 75 | |||
| 76 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 77 | _mm256_add_epi8(__m256i __a, __m256i __b) |
||
| 78 | {
|
||
| 79 | return (__m256i)((__v32qi)__a + (__v32qi)__b);
|
||
| 80 | } |
||
| 81 | |||
| 82 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 83 | _mm256_add_epi16(__m256i __a, __m256i __b) |
||
| 84 | {
|
||
| 85 | return (__m256i)((__v16hi)__a + (__v16hi)__b);
|
||
| 86 | } |
||
| 87 | |||
| 88 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 89 | _mm256_add_epi32(__m256i __a, __m256i __b) |
||
| 90 | {
|
||
| 91 | return (__m256i)((__v8si)__a + (__v8si)__b);
|
||
| 92 | } |
||
| 93 | |||
| 94 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 95 | _mm256_add_epi64(__m256i __a, __m256i __b) |
||
| 96 | {
|
||
| 97 | return __a + __b;
|
||
| 98 | } |
||
| 99 | |||
| 100 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 101 | _mm256_adds_epi8(__m256i __a, __m256i __b) |
||
| 102 | {
|
||
| 103 | return (__m256i)__builtin_ia32_paddsb256((__v32qi)__a, (__v32qi)__b);
|
||
| 104 | } |
||
| 105 | |||
| 106 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 107 | _mm256_adds_epi16(__m256i __a, __m256i __b) |
||
| 108 | {
|
||
| 109 | return (__m256i)__builtin_ia32_paddsw256((__v16hi)__a, (__v16hi)__b);
|
||
| 110 | } |
||
| 111 | |||
| 112 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 113 | _mm256_adds_epu8(__m256i __a, __m256i __b) |
||
| 114 | {
|
||
| 115 | return (__m256i)__builtin_ia32_paddusb256((__v32qi)__a, (__v32qi)__b);
|
||
| 116 | } |
||
| 117 | |||
| 118 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 119 | _mm256_adds_epu16(__m256i __a, __m256i __b) |
||
| 120 | {
|
||
| 121 | return (__m256i)__builtin_ia32_paddusw256((__v16hi)__a, (__v16hi)__b);
|
||
| 122 | } |
||
| 123 | |||
| 124 | #define _mm256_alignr_epi8(a, b, n) __extension__ ({ \
|
||
| 125 | __m256i __a = (a); \ |
||
| 126 | __m256i __b = (b); \ |
||
| 127 | (__m256i)__builtin_ia32_palignr256((__v32qi)__a, (__v32qi)__b, (n)); }) |
||
| 128 | |||
| 129 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 130 | _mm256_and_si256(__m256i __a, __m256i __b) |
||
| 131 | {
|
||
| 132 | return __a & __b;
|
||
| 133 | } |
||
| 134 | |||
| 135 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 136 | _mm256_andnot_si256(__m256i __a, __m256i __b) |
||
| 137 | {
|
||
| 138 | return ~__a & __b;
|
||
| 139 | } |
||
| 140 | |||
| 141 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 142 | _mm256_avg_epu8(__m256i __a, __m256i __b) |
||
| 143 | {
|
||
| 144 | return (__m256i)__builtin_ia32_pavgb256((__v32qi)__a, (__v32qi)__b);
|
||
| 145 | } |
||
| 146 | |||
| 147 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 148 | _mm256_avg_epu16(__m256i __a, __m256i __b) |
||
| 149 | {
|
||
| 150 | return (__m256i)__builtin_ia32_pavgw256((__v16hi)__a, (__v16hi)__b);
|
||
| 151 | } |
||
| 152 | |||
| 153 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 154 | _mm256_blendv_epi8(__m256i __V1, __m256i __V2, __m256i __M) |
||
| 155 | {
|
||
| 156 | return (__m256i)__builtin_ia32_pblendvb256((__v32qi)__V1, (__v32qi)__V2,
|
||
| 157 | (__v32qi)__M); |
||
| 158 | } |
||
| 159 | |||
| 160 | #define _mm256_blend_epi16(V1, V2, M) __extension__ ({ \
|
||
| 161 | __m256i __V1 = (V1); \ |
||
| 162 | __m256i __V2 = (V2); \ |
||
| 163 | (__m256d)__builtin_shufflevector((__v16hi)__V1, (__v16hi)__V2, \ |
||
| 164 | (((M) & 0x01) ? 16 : 0), \ |
||
| 165 | (((M) & 0x02) ? 17 : 1), \ |
||
| 166 | (((M) & 0x04) ? 18 : 2), \ |
||
| 167 | (((M) & 0x08) ? 19 : 3), \ |
||
| 168 | (((M) & 0x10) ? 20 : 4), \ |
||
| 169 | (((M) & 0x20) ? 21 : 5), \ |
||
| 170 | (((M) & 0x40) ? 22 : 6), \ |
||
| 171 | (((M) & 0x80) ? 23 : 7), \ |
||
| 172 | (((M) & 0x01) ? 24 : 8), \ |
||
| 173 | (((M) & 0x02) ? 25 : 9), \ |
||
| 174 | (((M) & 0x04) ? 26 : 10), \ |
||
| 175 | (((M) & 0x08) ? 27 : 11), \ |
||
| 176 | (((M) & 0x10) ? 28 : 12), \ |
||
| 177 | (((M) & 0x20) ? 29 : 13), \ |
||
| 178 | (((M) & 0x40) ? 30 : 14), \ |
||
| 179 | (((M) & 0x80) ? 31 : 15)); }) |
||
| 180 | |||
| 181 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 182 | _mm256_cmpeq_epi8(__m256i __a, __m256i __b) |
||
| 183 | {
|
||
| 184 | return (__m256i)((__v32qi)__a == (__v32qi)__b);
|
||
| 185 | } |
||
| 186 | |||
| 187 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 188 | _mm256_cmpeq_epi16(__m256i __a, __m256i __b) |
||
| 189 | {
|
||
| 190 | return (__m256i)((__v16hi)__a == (__v16hi)__b);
|
||
| 191 | } |
||
| 192 | |||
| 193 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 194 | _mm256_cmpeq_epi32(__m256i __a, __m256i __b) |
||
| 195 | {
|
||
| 196 | return (__m256i)((__v8si)__a == (__v8si)__b);
|
||
| 197 | } |
||
| 198 | |||
| 199 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 200 | _mm256_cmpeq_epi64(__m256i __a, __m256i __b) |
||
| 201 | {
|
||
| 202 | return (__m256i)(__a == __b);
|
||
| 203 | } |
||
| 204 | |||
| 205 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 206 | _mm256_cmpgt_epi8(__m256i __a, __m256i __b) |
||
| 207 | {
|
||
| 208 | return (__m256i)((__v32qi)__a > (__v32qi)__b);
|
||
| 209 | } |
||
| 210 | |||
| 211 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 212 | _mm256_cmpgt_epi16(__m256i __a, __m256i __b) |
||
| 213 | {
|
||
| 214 | return (__m256i)((__v16hi)__a > (__v16hi)__b);
|
||
| 215 | } |
||
| 216 | |||
| 217 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 218 | _mm256_cmpgt_epi32(__m256i __a, __m256i __b) |
||
| 219 | {
|
||
| 220 | return (__m256i)((__v8si)__a > (__v8si)__b);
|
||
| 221 | } |
||
| 222 | |||
| 223 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 224 | _mm256_cmpgt_epi64(__m256i __a, __m256i __b) |
||
| 225 | {
|
||
| 226 | return (__m256i)(__a > __b);
|
||
| 227 | } |
||
| 228 | |||
| 229 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 230 | _mm256_hadd_epi16(__m256i __a, __m256i __b) |
||
| 231 | {
|
||
| 232 | return (__m256i)__builtin_ia32_phaddw256((__v16hi)__a, (__v16hi)__b);
|
||
| 233 | } |
||
| 234 | |||
| 235 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 236 | _mm256_hadd_epi32(__m256i __a, __m256i __b) |
||
| 237 | {
|
||
| 238 | return (__m256i)__builtin_ia32_phaddd256((__v8si)__a, (__v8si)__b);
|
||
| 239 | } |
||
| 240 | |||
| 241 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 242 | _mm256_hadds_epi16(__m256i __a, __m256i __b) |
||
| 243 | {
|
||
| 244 | return (__m256i)__builtin_ia32_phaddsw256((__v16hi)__a, (__v16hi)__b);
|
||
| 245 | } |
||
| 246 | |||
| 247 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 248 | _mm256_hsub_epi16(__m256i __a, __m256i __b) |
||
| 249 | {
|
||
| 250 | return (__m256i)__builtin_ia32_phsubw256((__v16hi)__a, (__v16hi)__b);
|
||
| 251 | } |
||
| 252 | |||
| 253 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 254 | _mm256_hsub_epi32(__m256i __a, __m256i __b) |
||
| 255 | {
|
||
| 256 | return (__m256i)__builtin_ia32_phsubd256((__v8si)__a, (__v8si)__b);
|
||
| 257 | } |
||
| 258 | |||
| 259 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 260 | _mm256_hsubs_epi16(__m256i __a, __m256i __b) |
||
| 261 | {
|
||
| 262 | return (__m256i)__builtin_ia32_phsubsw256((__v16hi)__a, (__v16hi)__b);
|
||
| 263 | } |
||
| 264 | |||
| 265 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 266 | _mm256_maddubs_epi16(__m256i __a, __m256i __b) |
||
| 267 | {
|
||
| 268 | return (__m256i)__builtin_ia32_pmaddubsw256((__v32qi)__a, (__v32qi)__b);
|
||
| 269 | } |
||
| 270 | |||
| 271 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 272 | _mm256_madd_epi16(__m256i __a, __m256i __b) |
||
| 273 | {
|
||
| 274 | return (__m256i)__builtin_ia32_pmaddwd256((__v16hi)__a, (__v16hi)__b);
|
||
| 275 | } |
||
| 276 | |||
| 277 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 278 | _mm256_max_epi8(__m256i __a, __m256i __b) |
||
| 279 | {
|
||
| 280 | return (__m256i)__builtin_ia32_pmaxsb256((__v32qi)__a, (__v32qi)__b);
|
||
| 281 | } |
||
| 282 | |||
| 283 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 284 | _mm256_max_epi16(__m256i __a, __m256i __b) |
||
| 285 | {
|
||
| 286 | return (__m256i)__builtin_ia32_pmaxsw256((__v16hi)__a, (__v16hi)__b);
|
||
| 287 | } |
||
| 288 | |||
| 289 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 290 | _mm256_max_epi32(__m256i __a, __m256i __b) |
||
| 291 | {
|
||
| 292 | return (__m256i)__builtin_ia32_pmaxsd256((__v8si)__a, (__v8si)__b);
|
||
| 293 | } |
||
| 294 | |||
| 295 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 296 | _mm256_max_epu8(__m256i __a, __m256i __b) |
||
| 297 | {
|
||
| 298 | return (__m256i)__builtin_ia32_pmaxub256((__v32qi)__a, (__v32qi)__b);
|
||
| 299 | } |
||
| 300 | |||
| 301 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 302 | _mm256_max_epu16(__m256i __a, __m256i __b) |
||
| 303 | {
|
||
| 304 | return (__m256i)__builtin_ia32_pmaxuw256((__v16hi)__a, (__v16hi)__b);
|
||
| 305 | } |
||
| 306 | |||
| 307 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 308 | _mm256_max_epu32(__m256i __a, __m256i __b) |
||
| 309 | {
|
||
| 310 | return (__m256i)__builtin_ia32_pmaxud256((__v8si)__a, (__v8si)__b);
|
||
| 311 | } |
||
| 312 | |||
| 313 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 314 | _mm256_min_epi8(__m256i __a, __m256i __b) |
||
| 315 | {
|
||
| 316 | return (__m256i)__builtin_ia32_pminsb256((__v32qi)__a, (__v32qi)__b);
|
||
| 317 | } |
||
| 318 | |||
| 319 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 320 | _mm256_min_epi16(__m256i __a, __m256i __b) |
||
| 321 | {
|
||
| 322 | return (__m256i)__builtin_ia32_pminsw256((__v16hi)__a, (__v16hi)__b);
|
||
| 323 | } |
||
| 324 | |||
| 325 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 326 | _mm256_min_epi32(__m256i __a, __m256i __b) |
||
| 327 | {
|
||
| 328 | return (__m256i)__builtin_ia32_pminsd256((__v8si)__a, (__v8si)__b);
|
||
| 329 | } |
||
| 330 | |||
| 331 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 332 | _mm256_min_epu8(__m256i __a, __m256i __b) |
||
| 333 | {
|
||
| 334 | return (__m256i)__builtin_ia32_pminub256((__v32qi)__a, (__v32qi)__b);
|
||
| 335 | } |
||
| 336 | |||
| 337 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 338 | _mm256_min_epu16(__m256i __a, __m256i __b) |
||
| 339 | {
|
||
| 340 | return (__m256i)__builtin_ia32_pminuw256 ((__v16hi)__a, (__v16hi)__b);
|
||
| 341 | } |
||
| 342 | |||
| 343 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 344 | _mm256_min_epu32(__m256i __a, __m256i __b) |
||
| 345 | {
|
||
| 346 | return (__m256i)__builtin_ia32_pminud256((__v8si)__a, (__v8si)__b);
|
||
| 347 | } |
||
| 348 | |||
| 349 | static __inline__ int __attribute__((__always_inline__, __nodebug__)) |
||
| 350 | _mm256_movemask_epi8(__m256i __a) |
||
| 351 | {
|
||
| 352 | return __builtin_ia32_pmovmskb256((__v32qi)__a);
|
||
| 353 | } |
||
| 354 | |||
| 355 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 356 | _mm256_cvtepi8_epi16(__m128i __V) |
||
| 357 | {
|
||
| 358 | return (__m256i)__builtin_ia32_pmovsxbw256((__v16qi)__V);
|
||
| 359 | } |
||
| 360 | |||
| 361 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 362 | _mm256_cvtepi8_epi32(__m128i __V) |
||
| 363 | {
|
||
| 364 | return (__m256i)__builtin_ia32_pmovsxbd256((__v16qi)__V);
|
||
| 365 | } |
||
| 366 | |||
| 367 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 368 | _mm256_cvtepi8_epi64(__m128i __V) |
||
| 369 | {
|
||
| 370 | return (__m256i)__builtin_ia32_pmovsxbq256((__v16qi)__V);
|
||
| 371 | } |
||
| 372 | |||
| 373 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 374 | _mm256_cvtepi16_epi32(__m128i __V) |
||
| 375 | {
|
||
| 376 | return (__m256i)__builtin_ia32_pmovsxwd256((__v8hi)__V);
|
||
| 377 | } |
||
| 378 | |||
| 379 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 380 | _mm256_cvtepi16_epi64(__m128i __V) |
||
| 381 | {
|
||
| 382 | return (__m256i)__builtin_ia32_pmovsxwq256((__v8hi)__V);
|
||
| 383 | } |
||
| 384 | |||
| 385 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 386 | _mm256_cvtepi32_epi64(__m128i __V) |
||
| 387 | {
|
||
| 388 | return (__m256i)__builtin_ia32_pmovsxdq256((__v4si)__V);
|
||
| 389 | } |
||
| 390 | |||
| 391 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 392 | _mm256_cvtepu8_epi16(__m128i __V) |
||
| 393 | {
|
||
| 394 | return (__m256i)__builtin_ia32_pmovzxbw256((__v16qi)__V);
|
||
| 395 | } |
||
| 396 | |||
| 397 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 398 | _mm256_cvtepu8_epi32(__m128i __V) |
||
| 399 | {
|
||
| 400 | return (__m256i)__builtin_ia32_pmovzxbd256((__v16qi)__V);
|
||
| 401 | } |
||
| 402 | |||
| 403 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 404 | _mm256_cvtepu8_epi64(__m128i __V) |
||
| 405 | {
|
||
| 406 | return (__m256i)__builtin_ia32_pmovzxbq256((__v16qi)__V);
|
||
| 407 | } |
||
| 408 | |||
| 409 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 410 | _mm256_cvtepu16_epi32(__m128i __V) |
||
| 411 | {
|
||
| 412 | return (__m256i)__builtin_ia32_pmovzxwd256((__v8hi)__V);
|
||
| 413 | } |
||
| 414 | |||
| 415 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 416 | _mm256_cvtepu16_epi64(__m128i __V) |
||
| 417 | {
|
||
| 418 | return (__m256i)__builtin_ia32_pmovzxwq256((__v8hi)__V);
|
||
| 419 | } |
||
| 420 | |||
| 421 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 422 | _mm256_cvtepu32_epi64(__m128i __V) |
||
| 423 | {
|
||
| 424 | return (__m256i)__builtin_ia32_pmovzxdq256((__v4si)__V);
|
||
| 425 | } |
||
| 426 | |||
| 427 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 428 | _mm256_mul_epi32(__m256i __a, __m256i __b) |
||
| 429 | {
|
||
| 430 | return (__m256i)__builtin_ia32_pmuldq256((__v8si)__a, (__v8si)__b);
|
||
| 431 | } |
||
| 432 | |||
| 433 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 434 | _mm256_mulhrs_epi16(__m256i __a, __m256i __b) |
||
| 435 | {
|
||
| 436 | return (__m256i)__builtin_ia32_pmulhrsw256((__v16hi)__a, (__v16hi)__b);
|
||
| 437 | } |
||
| 438 | |||
| 439 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 440 | _mm256_mulhi_epu16(__m256i __a, __m256i __b) |
||
| 441 | {
|
||
| 442 | return (__m256i)__builtin_ia32_pmulhuw256((__v16hi)__a, (__v16hi)__b);
|
||
| 443 | } |
||
| 444 | |||
| 445 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 446 | _mm256_mulhi_epi16(__m256i __a, __m256i __b) |
||
| 447 | {
|
||
| 448 | return (__m256i)__builtin_ia32_pmulhw256((__v16hi)__a, (__v16hi)__b);
|
||
| 449 | } |
||
| 450 | |||
| 451 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 452 | _mm256_mullo_epi16(__m256i __a, __m256i __b) |
||
| 453 | {
|
||
| 454 | return (__m256i)((__v16hi)__a * (__v16hi)__b);
|
||
| 455 | } |
||
| 456 | |||
| 457 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 458 | _mm256_mullo_epi32 (__m256i __a, __m256i __b) |
||
| 459 | {
|
||
| 460 | return (__m256i)((__v8si)__a * (__v8si)__b);
|
||
| 461 | } |
||
| 462 | |||
| 463 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 464 | _mm256_mul_epu32(__m256i __a, __m256i __b) |
||
| 465 | {
|
||
| 466 | return __builtin_ia32_pmuludq256((__v8si)__a, (__v8si)__b);
|
||
| 467 | } |
||
| 468 | |||
| 469 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 470 | _mm256_or_si256(__m256i __a, __m256i __b) |
||
| 471 | {
|
||
| 472 | return __a | __b;
|
||
| 473 | } |
||
| 474 | |||
| 475 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 476 | _mm256_sad_epu8(__m256i __a, __m256i __b) |
||
| 477 | {
|
||
| 478 | return __builtin_ia32_psadbw256((__v32qi)__a, (__v32qi)__b);
|
||
| 479 | } |
||
| 480 | |||
| 481 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 482 | _mm256_shuffle_epi8(__m256i __a, __m256i __b) |
||
| 483 | {
|
||
| 484 | return (__m256i)__builtin_ia32_pshufb256((__v32qi)__a, (__v32qi)__b);
|
||
| 485 | } |
||
| 486 | |||
| 487 | #define _mm256_shuffle_epi32(a, imm) __extension__ ({ \
|
||
| 488 | __m256i __a = (a); \ |
||
| 489 | (__m256i)__builtin_shufflevector((__v8si)__a, (__v8si)_mm256_set1_epi32(0), \
|
||
| 490 | (imm) & 0x3, ((imm) & 0xc) >> 2, \ |
||
| 491 | ((imm) & 0x30) >> 4, ((imm) & 0xc0) >> 6, \ |
||
| 492 | 4 + (((imm) & 0x03) >> 0), \ |
||
| 493 | 4 + (((imm) & 0x0c) >> 2), \ |
||
| 494 | 4 + (((imm) & 0x30) >> 4), \ |
||
| 495 | 4 + (((imm) & 0xc0) >> 6)); }) |
||
| 496 | |||
| 497 | #define _mm256_shufflehi_epi16(a, imm) __extension__ ({ \
|
||
| 498 | __m256i __a = (a); \ |
||
| 499 | (__m256i)__builtin_shufflevector((__v16hi)__a, (__v16hi)_mm256_set1_epi16(0), \
|
||
| 500 | 0, 1, 2, 3, \ |
||
| 501 | 4 + (((imm) & 0x03) >> 0), \ |
||
| 502 | 4 + (((imm) & 0x0c) >> 2), \ |
||
| 503 | 4 + (((imm) & 0x30) >> 4), \ |
||
| 504 | 4 + (((imm) & 0xc0) >> 6), \ |
||
| 505 | 8, 9, 10, 11, \ |
||
| 506 | 12 + (((imm) & 0x03) >> 0), \ |
||
| 507 | 12 + (((imm) & 0x0c) >> 2), \ |
||
| 508 | 12 + (((imm) & 0x30) >> 4), \ |
||
| 509 | 12 + (((imm) & 0xc0) >> 6)); }) |
||
| 510 | |||
| 511 | #define _mm256_shufflelo_epi16(a, imm) __extension__ ({ \
|
||
| 512 | __m256i __a = (a); \ |
||
| 513 | (__m256i)__builtin_shufflevector((__v16hi)__a, (__v16hi)_mm256_set1_epi16(0), \
|
||
| 514 | (imm) & 0x3,((imm) & 0xc) >> 2, \ |
||
| 515 | ((imm) & 0x30) >> 4, ((imm) & 0xc0) >> 6, \ |
||
| 516 | 4, 5, 6, 7, \ |
||
| 517 | 8 + (((imm) & 0x03) >> 0), \ |
||
| 518 | 8 + (((imm) & 0x0c) >> 2), \ |
||
| 519 | 8 + (((imm) & 0x30) >> 4), \ |
||
| 520 | 8 + (((imm) & 0xc0) >> 6), \ |
||
| 521 | 12, 13, 14, 15); }) |
||
| 522 | |||
| 523 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 524 | _mm256_sign_epi8(__m256i __a, __m256i __b) |
||
| 525 | {
|
||
| 526 | return (__m256i)__builtin_ia32_psignb256((__v32qi)__a, (__v32qi)__b);
|
||
| 527 | } |
||
| 528 | |||
| 529 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 530 | _mm256_sign_epi16(__m256i __a, __m256i __b) |
||
| 531 | {
|
||
| 532 | return (__m256i)__builtin_ia32_psignw256((__v16hi)__a, (__v16hi)__b);
|
||
| 533 | } |
||
| 534 | |||
| 535 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 536 | _mm256_sign_epi32(__m256i __a, __m256i __b) |
||
| 537 | {
|
||
| 538 | return (__m256i)__builtin_ia32_psignd256((__v8si)__a, (__v8si)__b);
|
||
| 539 | } |
||
| 540 | |||
| 541 | #define _mm256_slli_si256(a, count) __extension__ ({ \
|
||
| 542 | __m256i __a = (a); \ |
||
| 543 | (__m256i)__builtin_ia32_pslldqi256(__a, (count)*8); })
|
||
| 544 | |||
| 545 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 546 | _mm256_slli_epi16(__m256i __a, int __count)
|
||
| 547 | {
|
||
| 548 | return (__m256i)__builtin_ia32_psllwi256((__v16hi)__a, __count);
|
||
| 549 | } |
||
| 550 | |||
| 551 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 552 | _mm256_sll_epi16(__m256i __a, __m128i __count) |
||
| 553 | {
|
||
| 554 | return (__m256i)__builtin_ia32_psllw256((__v16hi)__a, (__v8hi)__count);
|
||
| 555 | } |
||
| 556 | |||
| 557 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 558 | _mm256_slli_epi32(__m256i __a, int __count)
|
||
| 559 | {
|
||
| 560 | return (__m256i)__builtin_ia32_pslldi256((__v8si)__a, __count);
|
||
| 561 | } |
||
| 562 | |||
| 563 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 564 | _mm256_sll_epi32(__m256i __a, __m128i __count) |
||
| 565 | {
|
||
| 566 | return (__m256i)__builtin_ia32_pslld256((__v8si)__a, (__v4si)__count);
|
||
| 567 | } |
||
| 568 | |||
| 569 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 570 | _mm256_slli_epi64(__m256i __a, int __count)
|
||
| 571 | {
|
||
| 572 | return __builtin_ia32_psllqi256(__a, __count);
|
||
| 573 | } |
||
| 574 | |||
| 575 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 576 | _mm256_sll_epi64(__m256i __a, __m128i __count) |
||
| 577 | {
|
||
| 578 | return __builtin_ia32_psllq256(__a, __count);
|
||
| 579 | } |
||
| 580 | |||
| 581 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 582 | _mm256_srai_epi16(__m256i __a, int __count)
|
||
| 583 | {
|
||
| 584 | return (__m256i)__builtin_ia32_psrawi256((__v16hi)__a, __count);
|
||
| 585 | } |
||
| 586 | |||
| 587 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 588 | _mm256_sra_epi16(__m256i __a, __m128i __count) |
||
| 589 | {
|
||
| 590 | return (__m256i)__builtin_ia32_psraw256((__v16hi)__a, (__v8hi)__count);
|
||
| 591 | } |
||
| 592 | |||
| 593 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 594 | _mm256_srai_epi32(__m256i __a, int __count)
|
||
| 595 | {
|
||
| 596 | return (__m256i)__builtin_ia32_psradi256((__v8si)__a, __count);
|
||
| 597 | } |
||
| 598 | |||
| 599 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 600 | _mm256_sra_epi32(__m256i __a, __m128i __count) |
||
| 601 | {
|
||
| 602 | return (__m256i)__builtin_ia32_psrad256((__v8si)__a, (__v4si)__count);
|
||
| 603 | } |
||
| 604 | |||
| 605 | #define _mm256_srli_si256(a, count) __extension__ ({ \
|
||
| 606 | __m256i __a = (a); \ |
||
| 607 | (__m256i)__builtin_ia32_psrldqi256(__a, (count)*8); })
|
||
| 608 | |||
| 609 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 610 | _mm256_srli_epi16(__m256i __a, int __count)
|
||
| 611 | {
|
||
| 612 | return (__m256i)__builtin_ia32_psrlwi256((__v16hi)__a, __count);
|
||
| 613 | } |
||
| 614 | |||
| 615 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 616 | _mm256_srl_epi16(__m256i __a, __m128i __count) |
||
| 617 | {
|
||
| 618 | return (__m256i)__builtin_ia32_psrlw256((__v16hi)__a, (__v8hi)__count);
|
||
| 619 | } |
||
| 620 | |||
| 621 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 622 | _mm256_srli_epi32(__m256i __a, int __count)
|
||
| 623 | {
|
||
| 624 | return (__m256i)__builtin_ia32_psrldi256((__v8si)__a, __count);
|
||
| 625 | } |
||
| 626 | |||
| 627 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 628 | _mm256_srl_epi32(__m256i __a, __m128i __count) |
||
| 629 | {
|
||
| 630 | return (__m256i)__builtin_ia32_psrld256((__v8si)__a, (__v4si)__count);
|
||
| 631 | } |
||
| 632 | |||
| 633 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 634 | _mm256_srli_epi64(__m256i __a, int __count)
|
||
| 635 | {
|
||
| 636 | return __builtin_ia32_psrlqi256(__a, __count);
|
||
| 637 | } |
||
| 638 | |||
| 639 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 640 | _mm256_srl_epi64(__m256i __a, __m128i __count) |
||
| 641 | {
|
||
| 642 | return __builtin_ia32_psrlq256(__a, __count);
|
||
| 643 | } |
||
| 644 | |||
| 645 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 646 | _mm256_sub_epi8(__m256i __a, __m256i __b) |
||
| 647 | {
|
||
| 648 | return (__m256i)((__v32qi)__a - (__v32qi)__b);
|
||
| 649 | } |
||
| 650 | |||
| 651 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 652 | _mm256_sub_epi16(__m256i __a, __m256i __b) |
||
| 653 | {
|
||
| 654 | return (__m256i)((__v16hi)__a - (__v16hi)__b);
|
||
| 655 | } |
||
| 656 | |||
| 657 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 658 | _mm256_sub_epi32(__m256i __a, __m256i __b) |
||
| 659 | {
|
||
| 660 | return (__m256i)((__v8si)__a - (__v8si)__b);
|
||
| 661 | } |
||
| 662 | |||
| 663 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 664 | _mm256_sub_epi64(__m256i __a, __m256i __b) |
||
| 665 | {
|
||
| 666 | return __a - __b;
|
||
| 667 | } |
||
| 668 | |||
| 669 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 670 | _mm256_subs_epi8(__m256i __a, __m256i __b) |
||
| 671 | {
|
||
| 672 | return (__m256i)__builtin_ia32_psubsb256((__v32qi)__a, (__v32qi)__b);
|
||
| 673 | } |
||
| 674 | |||
| 675 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 676 | _mm256_subs_epi16(__m256i __a, __m256i __b) |
||
| 677 | {
|
||
| 678 | return (__m256i)__builtin_ia32_psubsw256((__v16hi)__a, (__v16hi)__b);
|
||
| 679 | } |
||
| 680 | |||
| 681 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 682 | _mm256_subs_epu8(__m256i __a, __m256i __b) |
||
| 683 | {
|
||
| 684 | return (__m256i)__builtin_ia32_psubusb256((__v32qi)__a, (__v32qi)__b);
|
||
| 685 | } |
||
| 686 | |||
| 687 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 688 | _mm256_subs_epu16(__m256i __a, __m256i __b) |
||
| 689 | {
|
||
| 690 | return (__m256i)__builtin_ia32_psubusw256((__v16hi)__a, (__v16hi)__b);
|
||
| 691 | } |
||
| 692 | |||
| 693 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 694 | _mm256_unpackhi_epi8(__m256i __a, __m256i __b) |
||
| 695 | {
|
||
| 696 | return (__m256i)__builtin_shufflevector((__v32qi)__a, (__v32qi)__b, 8, 32+8, 9, 32+9, 10, 32+10, 11, 32+11, 12, 32+12, 13, 32+13, 14, 32+14, 15, 32+15, 24, 32+24, 25, 32+25, 26, 32+26, 27, 32+27, 28, 32+28, 29, 32+29, 30, 32+30, 31, 32+31); |
||
| 697 | } |
||
| 698 | |||
| 699 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 700 | _mm256_unpackhi_epi16(__m256i __a, __m256i __b) |
||
| 701 | {
|
||
| 702 | return (__m256i)__builtin_shufflevector((__v16hi)__a, (__v16hi)__b, 4, 16+4, 5, 16+5, 6, 16+6, 7, 16+7, 12, 16+12, 13, 16+13, 14, 16+14, 15, 16+15); |
||
| 703 | } |
||
| 704 | |||
| 705 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 706 | _mm256_unpackhi_epi32(__m256i __a, __m256i __b) |
||
| 707 | {
|
||
| 708 | return (__m256i)__builtin_shufflevector((__v8si)__a, (__v8si)__b, 2, 8+2, 3, 8+3, 6, 8+6, 7, 8+7); |
||
| 709 | } |
||
| 710 | |||
| 711 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 712 | _mm256_unpackhi_epi64(__m256i __a, __m256i __b) |
||
| 713 | {
|
||
| 714 | return (__m256i)__builtin_shufflevector(__a, __b, 1, 4+1, 3, 4+3); |
||
| 715 | } |
||
| 716 | |||
| 717 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 718 | _mm256_unpacklo_epi8(__m256i __a, __m256i __b) |
||
| 719 | {
|
||
| 720 | return (__m256i)__builtin_shufflevector((__v32qi)__a, (__v32qi)__b, 0, 32+0, 1, 32+1, 2, 32+2, 3, 32+3, 4, 32+4, 5, 32+5, 6, 32+6, 7, 32+7, 16, 32+16, 17, 32+17, 18, 32+18, 19, 32+19, 20, 32+20, 21, 32+21, 22, 32+22, 23, 32+23); |
||
| 721 | } |
||
| 722 | |||
| 723 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 724 | _mm256_unpacklo_epi16(__m256i __a, __m256i __b) |
||
| 725 | {
|
||
| 726 | return (__m256i)__builtin_shufflevector((__v16hi)__a, (__v16hi)__b, 0, 16+0, 1, 16+1, 2, 16+2, 3, 16+3, 8, 16+8, 9, 16+9, 10, 16+10, 11, 16+11); |
||
| 727 | } |
||
| 728 | |||
| 729 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 730 | _mm256_unpacklo_epi32(__m256i __a, __m256i __b) |
||
| 731 | {
|
||
| 732 | return (__m256i)__builtin_shufflevector((__v8si)__a, (__v8si)__b, 0, 8+0, 1, 8+1, 4, 8+4, 5, 8+5); |
||
| 733 | } |
||
| 734 | |||
| 735 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 736 | _mm256_unpacklo_epi64(__m256i __a, __m256i __b) |
||
| 737 | {
|
||
| 738 | return (__m256i)__builtin_shufflevector(__a, __b, 0, 4+0, 2, 4+2); |
||
| 739 | } |
||
| 740 | |||
| 741 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 742 | _mm256_xor_si256(__m256i __a, __m256i __b) |
||
| 743 | {
|
||
| 744 | return __a ^ __b;
|
||
| 745 | } |
||
| 746 | |||
| 747 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 748 | _mm256_stream_load_si256(__m256i *__V) |
||
| 749 | {
|
||
| 750 | return (__m256i)__builtin_ia32_movntdqa256((__v4di *)__V);
|
||
| 751 | } |
||
| 752 | |||
| 753 | static __inline__ __m128 __attribute__((__always_inline__, __nodebug__))
|
||
| 754 | _mm_broadcastss_ps(__m128 __X) |
||
| 755 | {
|
||
| 756 | return (__m128)__builtin_ia32_vbroadcastss_ps((__v4sf)__X);
|
||
| 757 | } |
||
| 758 | |||
| 759 | static __inline__ __m256 __attribute__((__always_inline__, __nodebug__))
|
||
| 760 | _mm256_broadcastss_ps(__m128 __X) |
||
| 761 | {
|
||
| 762 | return (__m256)__builtin_ia32_vbroadcastss_ps256((__v4sf)__X);
|
||
| 763 | } |
||
| 764 | |||
| 765 | static __inline__ __m256d __attribute__((__always_inline__, __nodebug__))
|
||
| 766 | _mm256_broadcastsd_pd(__m128d __X) |
||
| 767 | {
|
||
| 768 | return (__m256d)__builtin_ia32_vbroadcastsd_pd256((__v2df)__X);
|
||
| 769 | } |
||
| 770 | |||
| 771 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 772 | _mm256_broadcastsi128_si256(__m128i __X) |
||
| 773 | {
|
||
| 774 | return (__m256i)__builtin_ia32_vbroadcastsi256(__X);
|
||
| 775 | } |
||
| 776 | |||
| 777 | #define _mm_blend_epi32(V1, V2, M) __extension__ ({ \
|
||
| 778 | __m128i __V1 = (V1); \ |
||
| 779 | __m128i __V2 = (V2); \ |
||
| 780 | (__m128i)__builtin_shufflevector((__v4si)__V1, (__v4si)__V2, \ |
||
| 781 | (((M) & 0x01) ? 4 : 0), \ |
||
| 782 | (((M) & 0x02) ? 5 : 1), \ |
||
| 783 | (((M) & 0x04) ? 6 : 2), \ |
||
| 784 | (((M) & 0x08) ? 7 : 3)); }) |
||
| 785 | |||
| 786 | #define _mm256_blend_epi32(V1, V2, M) __extension__ ({ \
|
||
| 787 | __m256i __V1 = (V1); \ |
||
| 788 | __m256i __V2 = (V2); \ |
||
| 789 | (__m256i)__builtin_shufflevector((__v8si)__V1, (__v8si)__V2, \ |
||
| 790 | (((M) & 0x01) ? 8 : 0), \ |
||
| 791 | (((M) & 0x02) ? 9 : 1), \ |
||
| 792 | (((M) & 0x04) ? 10 : 2), \ |
||
| 793 | (((M) & 0x08) ? 11 : 3), \ |
||
| 794 | (((M) & 0x10) ? 12 : 4), \ |
||
| 795 | (((M) & 0x20) ? 13 : 5), \ |
||
| 796 | (((M) & 0x40) ? 14 : 6), \ |
||
| 797 | (((M) & 0x80) ? 15 : 7)); }) |
||
| 798 | |||
| 799 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 800 | _mm256_broadcastb_epi8(__m128i __X) |
||
| 801 | {
|
||
| 802 | return (__m256i)__builtin_ia32_pbroadcastb256((__v16qi)__X);
|
||
| 803 | } |
||
| 804 | |||
| 805 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 806 | _mm256_broadcastw_epi16(__m128i __X) |
||
| 807 | {
|
||
| 808 | return (__m256i)__builtin_ia32_pbroadcastw256((__v8hi)__X);
|
||
| 809 | } |
||
| 810 | |||
| 811 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 812 | _mm256_broadcastd_epi32(__m128i __X) |
||
| 813 | {
|
||
| 814 | return (__m256i)__builtin_ia32_pbroadcastd256((__v4si)__X);
|
||
| 815 | } |
||
| 816 | |||
| 817 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 818 | _mm256_broadcastq_epi64(__m128i __X) |
||
| 819 | {
|
||
| 820 | return (__m256i)__builtin_ia32_pbroadcastq256(__X);
|
||
| 821 | } |
||
| 822 | |||
| 823 | static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
|
||
| 824 | _mm_broadcastb_epi8(__m128i __X) |
||
| 825 | {
|
||
| 826 | return (__m128i)__builtin_ia32_pbroadcastb128((__v16qi)__X);
|
||
| 827 | } |
||
| 828 | |||
| 829 | static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
|
||
| 830 | _mm_broadcastw_epi16(__m128i __X) |
||
| 831 | {
|
||
| 832 | return (__m128i)__builtin_ia32_pbroadcastw128((__v8hi)__X);
|
||
| 833 | } |
||
| 834 | |||
| 835 | |||
| 836 | static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
|
||
| 837 | _mm_broadcastd_epi32(__m128i __X) |
||
| 838 | {
|
||
| 839 | return (__m128i)__builtin_ia32_pbroadcastd128((__v4si)__X);
|
||
| 840 | } |
||
| 841 | |||
| 842 | static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
|
||
| 843 | _mm_broadcastq_epi64(__m128i __X) |
||
| 844 | {
|
||
| 845 | return (__m128i)__builtin_ia32_pbroadcastq128(__X);
|
||
| 846 | } |
||
| 847 | |||
| 848 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 849 | _mm256_permutevar8x32_epi32(__m256i __a, __m256i __b) |
||
| 850 | {
|
||
| 851 | return (__m256i)__builtin_ia32_permvarsi256((__v8si)__a, (__v8si)__b);
|
||
| 852 | } |
||
| 853 | |||
| 854 | #define _mm256_permute4x64_pd(V, M) __extension__ ({ \
|
||
| 855 | __m256d __V = (V); \ |
||
| 856 | (__m256d)__builtin_shufflevector((__v4df)__V, (__v4df) _mm256_setzero_pd(), \ |
||
| 857 | (M) & 0x3, ((M) & 0xc) >> 2, \ |
||
| 858 | ((M) & 0x30) >> 4, ((M) & 0xc0) >> 6); }) |
||
| 859 | |||
| 860 | static __inline__ __m256 __attribute__((__always_inline__, __nodebug__))
|
||
| 861 | _mm256_permutevar8x32_ps(__m256 __a, __m256 __b) |
||
| 862 | {
|
||
| 863 | return (__m256)__builtin_ia32_permvarsf256((__v8sf)__a, (__v8sf)__b);
|
||
| 864 | } |
||
| 865 | |||
| 866 | #define _mm256_permute4x64_epi64(V, M) __extension__ ({ \
|
||
| 867 | __m256i __V = (V); \ |
||
| 868 | (__m256i)__builtin_shufflevector((__v4di)__V, (__v4di) _mm256_setzero_si256(), \ |
||
| 869 | (M) & 0x3, ((M) & 0xc) >> 2, \ |
||
| 870 | ((M) & 0x30) >> 4, ((M) & 0xc0) >> 6); }) |
||
| 871 | |||
| 872 | #define _mm256_permute2x128_si256(V1, V2, M) __extension__ ({ \
|
||
| 873 | __m256i __V1 = (V1); \ |
||
| 874 | __m256i __V2 = (V2); \ |
||
| 875 | (__m256i)__builtin_ia32_permti256(__V1, __V2, (M)); }) |
||
| 876 | |||
| 877 | #define _mm256_extracti128_si256(A, O) __extension__ ({ \
|
||
| 878 | __m256i __A = (A); \ |
||
| 879 | (__m128i)__builtin_ia32_extract128i256(__A, (O)); }) |
||
| 880 | |||
| 881 | #define _mm256_inserti128_si256(V1, V2, O) __extension__ ({ \
|
||
| 882 | __m256i __V1 = (V1); \ |
||
| 883 | __m128i __V2 = (V2); \ |
||
| 884 | (__m256i)__builtin_ia32_insert128i256(__V1, __V2, (O)); }) |
||
| 885 | |||
| 886 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 887 | _mm256_maskload_epi32(int const *__X, __m256i __M) |
||
| 888 | {
|
||
| 889 | return (__m256i)__builtin_ia32_maskloadd256((const __v8si *)__X, (__v8si)__M); |
||
| 890 | } |
||
| 891 | |||
| 892 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 893 | _mm256_maskload_epi64(long long const *__X, __m256i __M) |
||
| 894 | {
|
||
| 895 | return (__m256i)__builtin_ia32_maskloadq256((const __v4di *)__X, __M); |
||
| 896 | } |
||
| 897 | |||
| 898 | static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
|
||
| 899 | _mm_maskload_epi32(int const *__X, __m128i __M) |
||
| 900 | {
|
||
| 901 | return (__m128i)__builtin_ia32_maskloadd((const __v4si *)__X, (__v4si)__M); |
||
| 902 | } |
||
| 903 | |||
| 904 | static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
|
||
| 905 | _mm_maskload_epi64(long long const *__X, __m128i __M) |
||
| 906 | {
|
||
| 907 | return (__m128i)__builtin_ia32_maskloadq((const __v2di *)__X, (__v2di)__M); |
||
| 908 | } |
||
| 909 | |||
| 910 | static __inline__ void __attribute__((__always_inline__, __nodebug__)) |
||
| 911 | _mm256_maskstore_epi32(int *__X, __m256i __M, __m256i __Y)
|
||
| 912 | {
|
||
| 913 | __builtin_ia32_maskstored256((__v8si *)__X, (__v8si)__M, (__v8si)__Y); |
||
| 914 | } |
||
| 915 | |||
| 916 | static __inline__ void __attribute__((__always_inline__, __nodebug__)) |
||
| 917 | _mm256_maskstore_epi64(long long *__X, __m256i __M, __m256i __Y) |
||
| 918 | {
|
||
| 919 | __builtin_ia32_maskstoreq256((__v4di *)__X, __M, __Y); |
||
| 920 | } |
||
| 921 | |||
| 922 | static __inline__ void __attribute__((__always_inline__, __nodebug__)) |
||
| 923 | _mm_maskstore_epi32(int *__X, __m128i __M, __m128i __Y)
|
||
| 924 | {
|
||
| 925 | __builtin_ia32_maskstored((__v4si *)__X, (__v4si)__M, (__v4si)__Y); |
||
| 926 | } |
||
| 927 | |||
| 928 | static __inline__ void __attribute__((__always_inline__, __nodebug__)) |
||
| 929 | _mm_maskstore_epi64(long long *__X, __m128i __M, __m128i __Y) |
||
| 930 | {
|
||
| 931 | __builtin_ia32_maskstoreq(( __v2di *)__X, __M, __Y); |
||
| 932 | } |
||
| 933 | |||
| 934 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 935 | _mm256_sllv_epi32(__m256i __X, __m256i __Y) |
||
| 936 | {
|
||
| 937 | return (__m256i)__builtin_ia32_psllv8si((__v8si)__X, (__v8si)__Y);
|
||
| 938 | } |
||
| 939 | |||
| 940 | static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
|
||
| 941 | _mm_sllv_epi32(__m128i __X, __m128i __Y) |
||
| 942 | {
|
||
| 943 | return (__m128i)__builtin_ia32_psllv4si((__v4si)__X, (__v4si)__Y);
|
||
| 944 | } |
||
| 945 | |||
| 946 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 947 | _mm256_sllv_epi64(__m256i __X, __m256i __Y) |
||
| 948 | {
|
||
| 949 | return (__m256i)__builtin_ia32_psllv4di(__X, __Y);
|
||
| 950 | } |
||
| 951 | |||
| 952 | static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
|
||
| 953 | _mm_sllv_epi64(__m128i __X, __m128i __Y) |
||
| 954 | {
|
||
| 955 | return (__m128i)__builtin_ia32_psllv2di(__X, __Y);
|
||
| 956 | } |
||
| 957 | |||
| 958 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 959 | _mm256_srav_epi32(__m256i __X, __m256i __Y) |
||
| 960 | {
|
||
| 961 | return (__m256i)__builtin_ia32_psrav8si((__v8si)__X, (__v8si)__Y);
|
||
| 962 | } |
||
| 963 | |||
| 964 | static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
|
||
| 965 | _mm_srav_epi32(__m128i __X, __m128i __Y) |
||
| 966 | {
|
||
| 967 | return (__m128i)__builtin_ia32_psrav4si((__v4si)__X, (__v4si)__Y);
|
||
| 968 | } |
||
| 969 | |||
| 970 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 971 | _mm256_srlv_epi32(__m256i __X, __m256i __Y) |
||
| 972 | {
|
||
| 973 | return (__m256i)__builtin_ia32_psrlv8si((__v8si)__X, (__v8si)__Y);
|
||
| 974 | } |
||
| 975 | |||
| 976 | static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
|
||
| 977 | _mm_srlv_epi32(__m128i __X, __m128i __Y) |
||
| 978 | {
|
||
| 979 | return (__m128i)__builtin_ia32_psrlv4si((__v4si)__X, (__v4si)__Y);
|
||
| 980 | } |
||
| 981 | |||
| 982 | static __inline__ __m256i __attribute__((__always_inline__, __nodebug__))
|
||
| 983 | _mm256_srlv_epi64(__m256i __X, __m256i __Y) |
||
| 984 | {
|
||
| 985 | return (__m256i)__builtin_ia32_psrlv4di(__X, __Y);
|
||
| 986 | } |
||
| 987 | |||
| 988 | static __inline__ __m128i __attribute__((__always_inline__, __nodebug__))
|
||
| 989 | _mm_srlv_epi64(__m128i __X, __m128i __Y) |
||
| 990 | {
|
||
| 991 | return (__m128i)__builtin_ia32_psrlv2di(__X, __Y);
|
||
| 992 | } |
||
| 993 | |||
| 994 | #define _mm_mask_i32gather_pd(a, m, i, mask, s) __extension__ ({ \
|
||
| 995 | __m128d __a = (a); \ |
||
| 996 | double const *__m = (m); \ |
||
| 997 | __m128i __i = (i); \ |
||
| 998 | __m128d __mask = (mask); \ |
||
| 999 | (__m128d)__builtin_ia32_gatherd_pd((__v2df)__a, (const __v2df *)__m, \
|
||
| 1000 | (__v4si)__i, (__v2df)__mask, (s)); }) |
||
| 1001 | |||
| 1002 | #define _mm256_mask_i32gather_pd(a, m, i, mask, s) __extension__ ({ \
|
||
| 1003 | __m256d __a = (a); \ |
||
| 1004 | double const *__m = (m); \ |
||
| 1005 | __m128i __i = (i); \ |
||
| 1006 | __m256d __mask = (mask); \ |
||
| 1007 | (__m256d)__builtin_ia32_gatherd_pd256((__v4df)__a, (const __v4df *)__m, \
|
||
| 1008 | (__v4si)__i, (__v4df)__mask, (s)); }) |
||
| 1009 | |||
| 1010 | #define _mm_mask_i64gather_pd(a, m, i, mask, s) __extension__ ({ \
|
||
| 1011 | __m128d __a = (a); \ |
||
| 1012 | double const *__m = (m); \ |
||
| 1013 | __m128i __i = (i); \ |
||
| 1014 | __m128d __mask = (mask); \ |
||
| 1015 | (__m128d)__builtin_ia32_gatherq_pd((__v2df)__a, (const __v2df *)__m, \
|
||
| 1016 | (__v2di)__i, (__v2df)__mask, (s)); }) |
||
| 1017 | |||
| 1018 | #define _mm256_mask_i64gather_pd(a, m, i, mask, s) __extension__ ({ \
|
||
| 1019 | __m256d __a = (a); \ |
||
| 1020 | double const *__m = (m); \ |
||
| 1021 | __m256i __i = (i); \ |
||
| 1022 | __m256d __mask = (mask); \ |
||
| 1023 | (__m256d)__builtin_ia32_gatherq_pd256((__v4df)__a, (const __v4df *)__m, \
|
||
| 1024 | (__v4di)__i, (__v4df)__mask, (s)); }) |
||
| 1025 | |||
| 1026 | #define _mm_mask_i32gather_ps(a, m, i, mask, s) __extension__ ({ \
|
||
| 1027 | __m128 __a = (a); \ |
||
| 1028 | float const *__m = (m); \ |
||
| 1029 | __m128i __i = (i); \ |
||
| 1030 | __m128 __mask = (mask); \ |
||
| 1031 | (__m128)__builtin_ia32_gatherd_ps((__v4sf)__a, (const __v4sf *)__m, \
|
||
| 1032 | (__v4si)__i, (__v4sf)__mask, (s)); }) |
||
| 1033 | |||
| 1034 | #define _mm256_mask_i32gather_ps(a, m, i, mask, s) __extension__ ({ \
|
||
| 1035 | __m256 __a = (a); \ |
||
| 1036 | float const *__m = (m); \ |
||
| 1037 | __m256i __i = (i); \ |
||
| 1038 | __m256 __mask = (mask); \ |
||
| 1039 | (__m256)__builtin_ia32_gatherd_ps256((__v8sf)__a, (const __v8sf *)__m, \
|
||
| 1040 | (__v8si)__i, (__v8sf)__mask, (s)); }) |
||
| 1041 | |||
| 1042 | #define _mm_mask_i64gather_ps(a, m, i, mask, s) __extension__ ({ \
|
||
| 1043 | __m128 __a = (a); \ |
||
| 1044 | float const *__m = (m); \ |
||
| 1045 | __m128i __i = (i); \ |
||
| 1046 | __m128 __mask = (mask); \ |
||
| 1047 | (__m128)__builtin_ia32_gatherq_ps((__v4sf)__a, (const __v4sf *)__m, \
|
||
| 1048 | (__v2di)__i, (__v4sf)__mask, (s)); }) |
||
| 1049 | |||
| 1050 | #define _mm256_mask_i64gather_ps(a, m, i, mask, s) __extension__ ({ \
|
||
| 1051 | __m128 __a = (a); \ |
||
| 1052 | float const *__m = (m); \ |
||
| 1053 | __m256i __i = (i); \ |
||
| 1054 | __m128 __mask = (mask); \ |
||
| 1055 | (__m128)__builtin_ia32_gatherq_ps256((__v4sf)__a, (const __v4sf *)__m, \
|
||
| 1056 | (__v4di)__i, (__v4sf)__mask, (s)); }) |
||
| 1057 | |||
| 1058 | #define _mm_mask_i32gather_epi32(a, m, i, mask, s) __extension__ ({ \
|
||
| 1059 | __m128i __a = (a); \ |
||
| 1060 | int const *__m = (m); \ |
||
| 1061 | __m128i __i = (i); \ |
||
| 1062 | __m128i __mask = (mask); \ |
||
| 1063 | (__m128i)__builtin_ia32_gatherd_d((__v4si)__a, (const __v4si *)__m, \
|
||
| 1064 | (__v4si)__i, (__v4si)__mask, (s)); }) |
||
| 1065 | |||
| 1066 | #define _mm256_mask_i32gather_epi32(a, m, i, mask, s) __extension__ ({ \
|
||
| 1067 | __m256i __a = (a); \ |
||
| 1068 | int const *__m = (m); \ |
||
| 1069 | __m256i __i = (i); \ |
||
| 1070 | __m256i __mask = (mask); \ |
||
| 1071 | (__m256i)__builtin_ia32_gatherd_d256((__v8si)__a, (const __v8si *)__m, \
|
||
| 1072 | (__v8si)__i, (__v8si)__mask, (s)); }) |
||
| 1073 | |||
| 1074 | #define _mm_mask_i64gather_epi32(a, m, i, mask, s) __extension__ ({ \
|
||
| 1075 | __m128i __a = (a); \ |
||
| 1076 | int const *__m = (m); \ |
||
| 1077 | __m128i __i = (i); \ |
||
| 1078 | __m128i __mask = (mask); \ |
||
| 1079 | (__m128i)__builtin_ia32_gatherq_d((__v4si)__a, (const __v4si *)__m, \
|
||
| 1080 | (__v2di)__i, (__v4si)__mask, (s)); }) |
||
| 1081 | |||
| 1082 | #define _mm256_mask_i64gather_epi32(a, m, i, mask, s) __extension__ ({ \
|
||
| 1083 | __m128i __a = (a); \ |
||
| 1084 | int const *__m = (m); \ |
||
| 1085 | __m256i __i = (i); \ |
||
| 1086 | __m128i __mask = (mask); \ |
||
| 1087 | (__m128i)__builtin_ia32_gatherq_d256((__v4si)__a, (const __v4si *)__m, \
|
||
| 1088 | (__v4di)__i, (__v4si)__mask, (s)); }) |
||
| 1089 | |||
| 1090 | #define _mm_mask_i32gather_epi64(a, m, i, mask, s) __extension__ ({ \
|
||
| 1091 | __m128i __a = (a); \ |
||
| 1092 | long long const *__m = (m); \ |
||
| 1093 | __m128i __i = (i); \ |
||
| 1094 | __m128i __mask = (mask); \ |
||
| 1095 | (__m128i)__builtin_ia32_gatherd_q((__v2di)__a, (const __v2di *)__m, \
|
||
| 1096 | (__v4si)__i, (__v2di)__mask, (s)); }) |
||
| 1097 | |||
| 1098 | #define _mm256_mask_i32gather_epi64(a, m, i, mask, s) __extension__ ({ \
|
||
| 1099 | __m256i __a = (a); \ |
||
| 1100 | long long const *__m = (m); \ |
||
| 1101 | __m128i __i = (i); \ |
||
| 1102 | __m256i __mask = (mask); \ |
||
| 1103 | (__m256i)__builtin_ia32_gatherd_q256((__v4di)__a, (const __v4di *)__m, \
|
||
| 1104 | (__v4si)__i, (__v4di)__mask, (s)); }) |
||
| 1105 | |||
| 1106 | #define _mm_mask_i64gather_epi64(a, m, i, mask, s) __extension__ ({ \
|
||
| 1107 | __m128i __a = (a); \ |
||
| 1108 | long long const *__m = (m); \ |
||
| 1109 | __m128i __i = (i); \ |
||
| 1110 | __m128i __mask = (mask); \ |
||
| 1111 | (__m128i)__builtin_ia32_gatherq_q((__v2di)__a, (const __v2di *)__m, \
|
||
| 1112 | (__v2di)__i, (__v2di)__mask, (s)); }) |
||
| 1113 | |||
| 1114 | #define _mm256_mask_i64gather_epi64(a, m, i, mask, s) __extension__ ({ \
|
||
| 1115 | __m256i __a = (a); \ |
||
| 1116 | long long const *__m = (m); \ |
||
| 1117 | __m256i __i = (i); \ |
||
| 1118 | __m256i __mask = (mask); \ |
||
| 1119 | (__m256i)__builtin_ia32_gatherq_q256((__v4di)__a, (const __v4di *)__m, \
|
||
| 1120 | (__v4di)__i, (__v4di)__mask, (s)); }) |
||
| 1121 | |||
| 1122 | #define _mm_i32gather_pd(m, i, s) __extension__ ({ \
|
||
| 1123 | double const *__m = (m); \ |
||
| 1124 | __m128i __i = (i); \ |
||
| 1125 | (__m128d)__builtin_ia32_gatherd_pd((__v2df)_mm_setzero_pd(), \ |
||
| 1126 | (const __v2df *)__m, (__v4si)__i, \
|
||
| 1127 | (__v2df)_mm_set1_pd((double)(long long int)-1), (s)); }) |
||
| 1128 | |||
| 1129 | #define _mm256_i32gather_pd(m, i, s) __extension__ ({ \
|
||
| 1130 | double const *__m = (m); \ |
||
| 1131 | __m128i __i = (i); \ |
||
| 1132 | (__m256d)__builtin_ia32_gatherd_pd256((__v4df)_mm256_setzero_pd(), \ |
||
| 1133 | (const __v4df *)__m, (__v4si)__i, \
|
||
| 1134 | (__v4df)_mm256_set1_pd((double)(long long int)-1), (s)); }) |
||
| 1135 | |||
| 1136 | #define _mm_i64gather_pd(m, i, s) __extension__ ({ \
|
||
| 1137 | double const *__m = (m); \ |
||
| 1138 | __m128i __i = (i); \ |
||
| 1139 | (__m128d)__builtin_ia32_gatherq_pd((__v2df)_mm_setzero_pd(), \ |
||
| 1140 | (const __v2df *)__m, (__v2di)__i, \
|
||
| 1141 | (__v2df)_mm_set1_pd((double)(long long int)-1), (s)); }) |
||
| 1142 | |||
| 1143 | #define _mm256_i64gather_pd(m, i, s) __extension__ ({ \
|
||
| 1144 | double const *__m = (m); \ |
||
| 1145 | __m256i __i = (i); \ |
||
| 1146 | (__m256d)__builtin_ia32_gatherq_pd256((__v4df)_mm256_setzero_pd(), \ |
||
| 1147 | (const __v4df *)__m, (__v4di)__i, \
|
||
| 1148 | (__v4df)_mm256_set1_pd((double)(long long int)-1), (s)); }) |
||
| 1149 | |||
| 1150 | #define _mm_i32gather_ps(m, i, s) __extension__ ({ \
|
||
| 1151 | float const *__m = (m); \ |
||
| 1152 | __m128i __i = (i); \ |
||
| 1153 | (__m128)__builtin_ia32_gatherd_ps((__v4sf)_mm_setzero_ps(), \ |
||
| 1154 | (const __v4sf *)__m, (__v4si)__i, \
|
||
| 1155 | (__v4sf)_mm_set1_ps((float)(int)-1), (s)); }) |
||
| 1156 | |||
| 1157 | #define _mm256_i32gather_ps(m, i, s) __extension__ ({ \
|
||
| 1158 | float const *__m = (m); \ |
||
| 1159 | __m256i __i = (i); \ |
||
| 1160 | (__m256)__builtin_ia32_gatherd_ps256((__v8sf)_mm256_setzero_ps(), \ |
||
| 1161 | (const __v8sf *)__m, (__v8si)__i, \
|
||
| 1162 | (__v8sf)_mm256_set1_ps((float)(int)-1), (s)); }) |
||
| 1163 | |||
| 1164 | #define _mm_i64gather_ps(m, i, s) __extension__ ({ \
|
||
| 1165 | float const *__m = (m); \ |
||
| 1166 | __m128i __i = (i); \ |
||
| 1167 | (__m128)__builtin_ia32_gatherq_ps((__v4sf)_mm_setzero_ps(), \ |
||
| 1168 | (const __v4sf *)__m, (__v2di)__i, \
|
||
| 1169 | (__v4sf)_mm_set1_ps((float)(int)-1), (s)); }) |
||
| 1170 | |||
| 1171 | #define _mm256_i64gather_ps(m, i, s) __extension__ ({ \
|
||
| 1172 | float const *__m = (m); \ |
||
| 1173 | __m256i __i = (i); \ |
||
| 1174 | (__m128)__builtin_ia32_gatherq_ps256((__v4sf)_mm_setzero_ps(), \ |
||
| 1175 | (const __v4sf *)__m, (__v4di)__i, \
|
||
| 1176 | (__v4sf)_mm_set1_ps((float)(int)-1), (s)); }) |
||
| 1177 | |||
| 1178 | #define _mm_i32gather_epi32(m, i, s) __extension__ ({ \
|
||
| 1179 | int const *__m = (m); \ |
||
| 1180 | __m128i __i = (i); \ |
||
| 1181 | (__m128i)__builtin_ia32_gatherd_d((__v4si)_mm_setzero_si128(), \ |
||
| 1182 | (const __v4si *)__m, (__v4si)__i, \
|
||
| 1183 | (__v4si)_mm_set1_epi32(-1), (s)); })
|
||
| 1184 | |||
| 1185 | #define _mm256_i32gather_epi32(m, i, s) __extension__ ({ \
|
||
| 1186 | int const *__m = (m); \ |
||
| 1187 | __m256i __i = (i); \ |
||
| 1188 | (__m256i)__builtin_ia32_gatherd_d256((__v8si)_mm256_setzero_si256(), \ |
||
| 1189 | (const __v8si *)__m, (__v8si)__i, \
|
||
| 1190 | (__v8si)_mm256_set1_epi32(-1), (s)); })
|
||
| 1191 | |||
| 1192 | #define _mm_i64gather_epi32(m, i, s) __extension__ ({ \
|
||
| 1193 | int const *__m = (m); \ |
||
| 1194 | __m128i __i = (i); \ |
||
| 1195 | (__m128i)__builtin_ia32_gatherq_d((__v4si)_mm_setzero_si128(), \ |
||
| 1196 | (const __v4si *)__m, (__v2di)__i, \
|
||
| 1197 | (__v4si)_mm_set1_epi32(-1), (s)); })
|
||
| 1198 | |||
| 1199 | #define _mm256_i64gather_epi32(m, i, s) __extension__ ({ \
|
||
| 1200 | int const *__m = (m); \ |
||
| 1201 | __m256i __i = (i); \ |
||
| 1202 | (__m128i)__builtin_ia32_gatherq_d256((__v4si)_mm_setzero_si128(), \ |
||
| 1203 | (const __v4si *)__m, (__v4di)__i, \
|
||
| 1204 | (__v4si)_mm_set1_epi32(-1), (s)); })
|
||
| 1205 | |||
| 1206 | #define _mm_i32gather_epi64(m, i, s) __extension__ ({ \
|
||
| 1207 | long long const *__m = (m); \ |
||
| 1208 | __m128i __i = (i); \ |
||
| 1209 | (__m128i)__builtin_ia32_gatherd_q((__v2di)_mm_setzero_si128(), \ |
||
| 1210 | (const __v2di *)__m, (__v4si)__i, \
|
||
| 1211 | (__v2di)_mm_set1_epi64x(-1), (s)); })
|
||
| 1212 | |||
| 1213 | #define _mm256_i32gather_epi64(m, i, s) __extension__ ({ \
|
||
| 1214 | long long const *__m = (m); \ |
||
| 1215 | __m128i __i = (i); \ |
||
| 1216 | (__m256i)__builtin_ia32_gatherd_q256((__v4di)_mm256_setzero_si256(), \ |
||
| 1217 | (const __v4di *)__m, (__v4si)__i, \
|
||
| 1218 | (__v4di)_mm256_set1_epi64x(-1), (s)); })
|
||
| 1219 | |||
| 1220 | #define _mm_i64gather_epi64(m, i, s) __extension__ ({ \
|
||
| 1221 | long long const *__m = (m); \ |
||
| 1222 | __m128i __i = (i); \ |
||
| 1223 | (__m128i)__builtin_ia32_gatherq_q((__v2di)_mm_setzero_si128(), \ |
||
| 1224 | (const __v2di *)__m, (__v2di)__i, \
|
||
| 1225 | (__v2di)_mm_set1_epi64x(-1), (s)); })
|
||
| 1226 | |||
| 1227 | #define _mm256_i64gather_epi64(m, i, s) __extension__ ({ \
|
||
| 1228 | long long const *__m = (m); \ |
||
| 1229 | __m256i __i = (i); \ |
||
| 1230 | (__m256i)__builtin_ia32_gatherq_q256((__v4di)_mm256_setzero_si256(), \ |
||
| 1231 | (const __v4di *)__m, (__v4di)__i, \
|
||
| 1232 | (__v4di)_mm256_set1_epi64x(-1), (s)); })
|
||
| 1233 | |||
| 1234 | #endif /* __AVX2INTRIN_H */ |