immintrin.h revision 1.2
1 /* $NetBSD: immintrin.h,v 1.2 2024/07/15 13:51:10 riastradh Exp $ */ 2 3 /*- 4 * Copyright (c) 2020 The NetBSD Foundation, Inc. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 16 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS 17 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED 18 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 19 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS 20 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR 21 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF 22 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS 23 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN 24 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 25 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE 26 * POSSIBILITY OF SUCH DAMAGE. 27 */ 28 29 #ifndef _SYS_CRYPTO_ARCH_X86_IMMINTRIN_H 30 #define _SYS_CRYPTO_ARCH_X86_IMMINTRIN_H 31 32 #include <sys/types.h> 33 34 /* 35 * This kludgerous header file provides definitions for the Intel 36 * intrinsics that work with GCC and Clang, because <immintrin.h> is 37 * not available during the kernel build and arranging to make it 38 * available is complicated. Please fix this properly! 39 */ 40 41 #if defined(__GNUC__) && !defined(__clang__) 42 43 #define _INTRINSATTR \ 44 __attribute__((__gnu_inline__, __always_inline__, __artificial__)) 45 46 typedef short __m16 __attribute__((__vector_size__(2), __may_alias__)); 47 typedef short __m16_u 48 __attribute__((__vector_size__(2), __may_alias__, __aligned__(1))); 49 typedef int __m32 __attribute__((__vector_size__(4), __may_alias__)); 50 typedef int __m32_u 51 __attribute__((__vector_size__(4), __may_alias__, __aligned__(1))); 52 typedef int __m64 __attribute__((__vector_size__(8), __may_alias__)); 53 typedef int __m64_u 54 __attribute__((__vector_size__(8), __may_alias__, __aligned__(1))); 55 typedef float __m128 __attribute__((__vector_size__(16), __may_alias__)); 56 typedef long long __m128i __attribute__((__vector_size__(16), __may_alias__)); 57 typedef long long __m128i_u 58 __attribute__((__vector_size__(16), __may_alias__, __aligned__(1))); 59 typedef long long __v2di __attribute__((__vector_size__(16))); 60 typedef unsigned long long __v2du __attribute__((__vector_size__(16))); 61 typedef int __v4si __attribute__((__vector_size__(16))); 62 typedef unsigned __v4su __attribute__((__vector_size__(16))); 63 typedef float __v4sf __attribute__((__vector_size__(16))); 64 typedef short __v8hi __attribute__((__vector_size__(16))); 65 typedef char __v16qi __attribute__((__vector_size__(16))); 66 67 #elif defined(__clang__) 68 69 typedef float __m128 __attribute__((__vector_size__(16), __aligned__(16))); 70 typedef long long __m128i 71 __attribute__((__vector_size__(16), __aligned__(16))); 72 typedef long long __m128i_u 73 __attribute__((__vector_size__(16), __may_alias__, __aligned__(1))); 74 typedef long long __v2di __attribute__((__vector_size__(16))); 75 typedef unsigned long long __v2du __attribute__((__vector_size__(16))); 76 typedef int __v4si __attribute__((__vector_size__(16))); 77 typedef unsigned __v4su __attribute__((__vector_size__(16))); 78 typedef float __v4sf __attribute__((__vector_size__(16))); 79 typedef short __v8hi __attribute__((__vector_size__(16))); 80 typedef char __v16qi __attribute__((__vector_size__(16))); 81 82 #define _INTRINSATTR \ 83 __attribute__((__always_inline__, __nodebug__, __target__("sse2"), \ 84 __min_vector_width__(128))) 85 #define _PACKALIAS \ 86 __attribute__((__packed__, __may_alias__)) 87 88 #else 89 90 #error Please teach me how to do Intel intrinsics for your compiler! 91 92 #endif 93 94 #define _SSSE3_ATTR __attribute__((target("ssse3"))) 95 96 _INTRINSATTR 97 static __inline __m128i 98 _mm_add_epi32(__m128i __a, __m128i __b) 99 { 100 return (__m128i)((__v4su)__a + (__v4su)__b); 101 } 102 103 #if defined(__GNUC__) && !defined(__clang__) 104 #define _mm_alignr_epi8(hi,lo,bytes) \ 105 (__m128i)__builtin_ia32_palignr128((__v2di)(__m128i)(hi), \ 106 (__v2di)(__m128i)(lo), 8*(int)(bytes)) 107 #elif defined(__clang__) 108 #define _mm_alignr_epi8(hi,lo,bytes) \ 109 (__m128i)__builtin_ia32_palignr128((__v16qi)(__m128i)(hi), \ 110 (__v16qi)(__m128i)(lo), (int)(bytes)) 111 #endif 112 113 _INTRINSATTR 114 static __inline __m128 115 _mm_load1_ps(const float *__p) 116 { 117 return __extension__ (__m128)(__v4sf) { *__p, *__p, *__p, *__p }; 118 } 119 120 _INTRINSATTR 121 static __inline __m128i 122 _mm_loadu_si128(const __m128i_u *__p) 123 { 124 #if defined(__GNUC__) && !defined(__clang__) 125 return *__p; 126 #else 127 return ((const struct { __m128i_u __v; } _PACKALIAS *)__p)->__v; 128 #endif 129 } 130 131 _INTRINSATTR 132 static __inline __m128i 133 _mm_loadu_si32(const void *__p) 134 { 135 #if defined(__GNUC__) && !defined(__clang__) 136 int32_t __v = (*(__m32_u *)__p)[0]; 137 #else 138 int32_t __v = ((const struct { int32_t __v; } _PACKALIAS *)__p)->__v; 139 #endif 140 return __extension__ (__m128i)(__v4si){ __v, 0, 0, 0 }; 141 } 142 143 _INTRINSATTR 144 static __inline __m128i 145 _mm_loadu_si64(const void *__p) 146 { 147 #if defined(__GNUC__) && !defined(__clang__) 148 int64_t __v = (*(__m64_u *)__p)[0]; 149 #else 150 int64_t __v = ((const struct { int64_t __v; } _PACKALIAS *)__p)->__v; 151 #endif 152 return __extension__ (__m128i)(__v2di){ __v, 0 }; 153 } 154 155 _INTRINSATTR 156 static __inline __m128i 157 _mm_load_si128(const __m128i *__p) 158 { 159 return *__p; 160 } 161 162 _INTRINSATTR 163 static __inline __m128 164 _mm_movehl_ps(__m128 __v0, __m128 __v1) 165 { 166 #if defined(__GNUC__) && !defined(__clang__) 167 return (__m128)__builtin_ia32_movhlps((__v4sf)__v0, (__v4sf)__v1); 168 #elif defined(__clang__) 169 return __builtin_shufflevector((__v4sf)__v0, (__v4sf)__v1, 6,7,2,3); 170 #endif 171 } 172 173 _INTRINSATTR 174 static __inline __m128 175 _mm_movelh_ps(__m128 __v0, __m128 __v1) 176 { 177 #if defined(__GNUC__) && !defined(__clang__) 178 return (__m128)__builtin_ia32_movlhps((__v4sf)__v0, (__v4sf)__v1); 179 #elif defined(__clang__) 180 return __builtin_shufflevector((__v4sf)__v0, (__v4sf)__v1, 0,1,4,5); 181 #endif 182 } 183 184 _INTRINSATTR 185 static __inline __m128i 186 _mm_set1_epi16(int16_t __v) 187 { 188 return __extension__ (__m128i)(__v8hi){ 189 __v, __v, __v, __v, __v, __v, __v, __v 190 }; 191 } 192 193 _INTRINSATTR 194 static __inline __m128i 195 _mm_set1_epi32(int32_t __v) 196 { 197 return __extension__ (__m128i)(__v4si){ __v, __v, __v, __v }; 198 } 199 200 _INTRINSATTR 201 static __inline __m128i 202 _mm_set1_epi64x(int64_t __v) 203 { 204 return __extension__ (__m128i)(__v2di){ __v, __v }; 205 } 206 207 _INTRINSATTR 208 static __inline __m128i 209 _mm_set_epi32(int32_t __v3, int32_t __v2, int32_t __v1, int32_t __v0) 210 { 211 return __extension__ (__m128i)(__v4si){ __v0, __v1, __v2, __v3 }; 212 } 213 214 _INTRINSATTR 215 static __inline __m128i 216 _mm_set_epi64x(int64_t __v1, int64_t __v0) 217 { 218 return __extension__ (__m128i)(__v2di){ __v0, __v1 }; 219 } 220 221 _INTRINSATTR 222 static __inline __m128 223 _mm_setzero_ps(void) 224 { 225 return __extension__ (__m128){ 0, 0, 0, 0 }; 226 } 227 228 _INTRINSATTR 229 static __inline __m128i 230 _mm_setzero_si128(void) 231 { 232 return _mm_set1_epi64x(0); 233 } 234 235 _INTRINSATTR _SSSE3_ATTR 236 static __inline __m128i 237 _mm_shuffle_epi8(__m128i __vtbl, __m128i __vidx) 238 { 239 return (__m128i)__builtin_ia32_pshufb128((__v16qi)__vtbl, 240 (__v16qi)__vidx); 241 } 242 243 #define _mm_shuffle_epi32(v,m) \ 244 (__m128i)__builtin_ia32_pshufd((__v4si)(__m128i)(v), (int)(m)) 245 246 #define _mm_shuffle_ps(x,y,m) \ 247 (__m128)__builtin_ia32_shufps((__v4sf)(__m128)(x), \ 248 (__v4sf)(__m128)(y), (int)(m)) \ 249 250 _INTRINSATTR 251 static __inline __m128i 252 _mm_slli_epi32(__m128i __v, uint8_t __bits) 253 { 254 return (__m128i)__builtin_ia32_pslldi128((__v4si)__v, (int)__bits); 255 } 256 257 _INTRINSATTR 258 static __inline __m128i 259 _mm_slli_epi64(__m128i __v, uint8_t __bits) 260 { 261 return (__m128i)__builtin_ia32_psllqi128((__v2di)__v, (int)__bits); 262 } 263 264 #if defined(__GNUC__) && !defined(__clang__) 265 #define _mm_slli_si128(v,bytes) \ 266 (__m128i)__builtin_ia32_pslldqi128((__v2di)(__m128i)(v), \ 267 8*(int)(bytes)) 268 #elif defined(__clang__) 269 #define _mm_slli_si128(v,bytes) \ 270 (__m128i)__builtin_ia32_pslldqi128_byteshift((__v2di)(__m128i)(v), \ 271 (int)(bytes)) 272 #endif 273 274 _INTRINSATTR 275 static __inline __m128i 276 _mm_srli_epi32(__m128i __v, uint8_t __bits) 277 { 278 return (__m128i)__builtin_ia32_psrldi128((__v4si)__v, (int)__bits); 279 } 280 281 _INTRINSATTR 282 static __inline __m128i 283 _mm_srli_epi64(__m128i __v, uint8_t __bits) 284 { 285 return (__m128i)__builtin_ia32_psrlqi128((__v2di)__v, (int)__bits); 286 } 287 288 #if defined(__GNUC__) && !defined(__clang__) 289 #define _mm_srli_si128(v,bytes) \ 290 (__m128i)__builtin_ia32_psrldqi128((__m128i)(v), 8*(int)(bytes)) 291 #elif defined(__clang__) 292 #define _mm_srli_si128(v,bytes) \ 293 (__m128i)__builtin_ia32_psrldqi128_byteshift((__v2di)(__m128i)(v), \ 294 (int)(bytes)); 295 #endif 296 297 _INTRINSATTR 298 static __inline void 299 _mm_storeu_si128(__m128i_u *__p, __m128i __v) 300 { 301 #if defined(__GNUC__) && !defined(__clang__) 302 *__p = __v; 303 #else 304 ((struct { __m128i_u __v; } _PACKALIAS *)__p)->__v = __v; 305 #endif 306 } 307 308 _INTRINSATTR 309 static __inline void 310 _mm_storeu_si32(void *__p, __m128i __v) 311 { 312 #if defined(__GNUC__) && !defined(__clang__) 313 *(__m32_u *)__p = (__m32)((__v4si)__v)[0]; 314 #else 315 ((struct { int32_t __v; } _PACKALIAS *)__p)->__v = ((__v4si)__v)[0]; 316 #endif 317 } 318 319 _INTRINSATTR 320 static __inline void 321 _mm_storeu_si64(void *__p, __m128i __v) 322 { 323 #if defined(__GNUC__) && !defined(__clang__) 324 *(__m64_u *)__p = (__m64)((__v2di)__v)[0]; 325 #else 326 ((struct { int64_t __v; } _PACKALIAS *)__p)->__v = ((__v2di)__v)[0]; 327 #endif 328 } 329 330 _INTRINSATTR 331 static __inline void 332 _mm_store_si128(__m128i *__p, __m128i __v) 333 { 334 *__p = __v; 335 } 336 337 _INTRINSATTR 338 static __inline __m128i 339 _mm_sub_epi64(__m128i __x, __m128i __y) 340 { 341 return (__m128i)((__v2du)__x - (__v2du)__y); 342 } 343 344 _INTRINSATTR 345 static __inline __m128i 346 _mm_unpackhi_epi32(__m128i __lo, __m128i __hi) 347 { 348 #if defined(__GNUC__) && !defined(__clang__) 349 return (__m128i)__builtin_ia32_punpckhdq128((__v4si)__lo, 350 (__v4si)__hi); 351 #elif defined(__clang__) 352 return (__m128i)__builtin_shufflevector((__v4si)__lo, (__v4si)__hi, 353 2,6,3,7); 354 #endif 355 } 356 357 _INTRINSATTR 358 static __inline __m128i 359 _mm_unpacklo_epi32(__m128i __lo, __m128i __hi) 360 { 361 #if defined(__GNUC__) && !defined(__clang__) 362 return (__m128i)__builtin_ia32_punpckldq128((__v4si)__lo, 363 (__v4si)__hi); 364 #elif defined(__clang__) 365 return (__m128i)__builtin_shufflevector((__v4si)__lo, (__v4si)__hi, 366 0,4,1,5); 367 #endif 368 } 369 370 _INTRINSATTR 371 static __inline __m128i 372 _mm_unpacklo_epi64(__m128i __lo, __m128i __hi) 373 { 374 #if defined(__GNUC__) && !defined(__clang__) 375 return (__m128i)__builtin_ia32_punpcklqdq128((__v2di)__lo, 376 (__v2di)__hi); 377 #elif defined(__clang__) 378 return (__m128i)__builtin_shufflevector((__v2di)__lo, (__v2di)__hi, 379 0,2); 380 #endif 381 } 382 383 #endif /* _SYS_CRYPTO_ARCH_X86_IMMINTRIN_H */ 384
Indexes created Mon Sep 29 03:10:08 GMT 2025