1 /* $NetBSD: immintrin.h,v 1.4 2024/07/16 15:27:40 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 typedef char __v16qi_u 67 __attribute__((__vector_size__(16), __may_alias__, __aligned__(1))); 68 69 #elif defined(__clang__) 70 71 typedef float __m128 __attribute__((__vector_size__(16), __aligned__(16))); 72 typedef long long __m128i 73 __attribute__((__vector_size__(16), __aligned__(16))); 74 typedef long long __m128i_u 75 __attribute__((__vector_size__(16), __may_alias__, __aligned__(1))); 76 typedef long long __v2di __attribute__((__vector_size__(16))); 77 typedef unsigned long long __v2du __attribute__((__vector_size__(16))); 78 typedef int __v4si __attribute__((__vector_size__(16))); 79 typedef unsigned __v4su __attribute__((__vector_size__(16))); 80 typedef float __v4sf __attribute__((__vector_size__(16))); 81 typedef short __v8hi __attribute__((__vector_size__(16))); 82 typedef char __v16qi __attribute__((__vector_size__(16))); 83 84 #define _INTRINSATTR \ 85 __attribute__((__always_inline__, __nodebug__, __target__("sse2"), \ 86 __min_vector_width__(128))) 87 #define _PACKALIAS \ 88 __attribute__((__packed__, __may_alias__)) 89 90 #else 91 92 #error Please teach me how to do Intel intrinsics for your compiler! 93 94 #endif 95 96 #define _SSSE3_ATTR __attribute__((target("ssse3"))) 97 98 _INTRINSATTR 99 static __inline __m128i 100 _mm_add_epi32(__m128i __a, __m128i __b) 101 { 102 return (__m128i)((__v4su)__a + (__v4su)__b); 103 } 104 105 #if defined(__GNUC__) && !defined(__clang__) 106 #define _mm_alignr_epi8(hi,lo,bytes) \ 107 (__m128i)__builtin_ia32_palignr128((__v2di)(__m128i)(hi), \ 108 (__v2di)(__m128i)(lo), 8*(int)(bytes)) 109 #elif defined(__clang__) 110 #define _mm_alignr_epi8(hi,lo,bytes) \ 111 (__m128i)__builtin_ia32_palignr128((__v16qi)(__m128i)(hi), \ 112 (__v16qi)(__m128i)(lo), (int)(bytes)) 113 #endif 114 115 _INTRINSATTR 116 static __inline __m128 117 _mm_load1_ps(const float *__p) 118 { 119 return __extension__ (__m128)(__v4sf) { *__p, *__p, *__p, *__p }; 120 } 121 122 _INTRINSATTR 123 static __inline __m128i 124 _mm_loadu_si128(const __m128i_u *__p) 125 { 126 #if defined(__GNUC__) && !defined(__clang__) 127 return *__p; 128 #else 129 return ((const struct { __m128i_u __v; } _PACKALIAS *)__p)->__v; 130 #endif 131 } 132 133 _INTRINSATTR 134 static __inline __m128i 135 _mm_loadu_si32(const void *__p) 136 { 137 #if defined(__GNUC__) && !defined(__clang__) 138 int32_t __v = (*(const __m32_u *)__p)[0]; 139 #else 140 int32_t __v = ((const struct { int32_t __v; } _PACKALIAS *)__p)->__v; 141 #endif 142 return __extension__ (__m128i)(__v4si){ __v, 0, 0, 0 }; 143 } 144 145 _INTRINSATTR 146 static __inline __m128i 147 _mm_loadu_si64(const void *__p) 148 { 149 #if defined(__GNUC__) && !defined(__clang__) 150 int64_t __v = (int64_t)*(const __m64_u *)__p; 151 #else 152 int64_t __v = ((const struct { int64_t __v; } _PACKALIAS *)__p)->__v; 153 #endif 154 return __extension__ (__m128i)(__v2di){ __v, 0 }; 155 } 156 157 _INTRINSATTR 158 static __inline __m128i 159 _mm_load_si128(const __m128i *__p) 160 { 161 return *__p; 162 } 163 164 _INTRINSATTR 165 static __inline __m128 166 _mm_movehl_ps(__m128 __v0, __m128 __v1) 167 { 168 #if defined(__GNUC__) && !defined(__clang__) 169 return (__m128)__builtin_ia32_movhlps((__v4sf)__v0, (__v4sf)__v1); 170 #elif defined(__clang__) 171 return __builtin_shufflevector((__v4sf)__v0, (__v4sf)__v1, 6,7,2,3); 172 #endif 173 } 174 175 _INTRINSATTR 176 static __inline __m128 177 _mm_movelh_ps(__m128 __v0, __m128 __v1) 178 { 179 #if defined(__GNUC__) && !defined(__clang__) 180 return (__m128)__builtin_ia32_movlhps((__v4sf)__v0, (__v4sf)__v1); 181 #elif defined(__clang__) 182 return __builtin_shufflevector((__v4sf)__v0, (__v4sf)__v1, 0,1,4,5); 183 #endif 184 } 185 186 _INTRINSATTR 187 static __inline __m128i 188 _mm_set1_epi16(int16_t __v) 189 { 190 return __extension__ (__m128i)(__v8hi){ 191 __v, __v, __v, __v, __v, __v, __v, __v 192 }; 193 } 194 195 _INTRINSATTR 196 static __inline __m128i 197 _mm_set1_epi32(int32_t __v) 198 { 199 return __extension__ (__m128i)(__v4si){ __v, __v, __v, __v }; 200 } 201 202 _INTRINSATTR 203 static __inline __m128i 204 _mm_set1_epi64x(int64_t __v) 205 { 206 return __extension__ (__m128i)(__v2di){ __v, __v }; 207 } 208 209 _INTRINSATTR 210 static __inline __m128i 211 _mm_set_epi32(int32_t __v3, int32_t __v2, int32_t __v1, int32_t __v0) 212 { 213 return __extension__ (__m128i)(__v4si){ __v0, __v1, __v2, __v3 }; 214 } 215 216 _INTRINSATTR 217 static __inline __m128i 218 _mm_set_epi64x(int64_t __v1, int64_t __v0) 219 { 220 return __extension__ (__m128i)(__v2di){ __v0, __v1 }; 221 } 222 223 _INTRINSATTR 224 static __inline __m128 225 _mm_setzero_ps(void) 226 { 227 return __extension__ (__m128){ 0, 0, 0, 0 }; 228 } 229 230 _INTRINSATTR 231 static __inline __m128i 232 _mm_setzero_si128(void) 233 { 234 return _mm_set1_epi64x(0); 235 } 236 237 _INTRINSATTR _SSSE3_ATTR 238 static __inline __m128i 239 _mm_shuffle_epi8(__m128i __vtbl, __m128i __vidx) 240 { 241 return (__m128i)__builtin_ia32_pshufb128((__v16qi)__vtbl, 242 (__v16qi)__vidx); 243 } 244 245 #define _mm_shuffle_epi32(v,m) \ 246 (__m128i)__builtin_ia32_pshufd((__v4si)(__m128i)(v), (int)(m)) 247 248 #define _mm_shuffle_ps(x,y,m) \ 249 (__m128)__builtin_ia32_shufps((__v4sf)(__m128)(x), \ 250 (__v4sf)(__m128)(y), (int)(m)) \ 251 252 _INTRINSATTR 253 static __inline __m128i 254 _mm_slli_epi32(__m128i __v, uint8_t __bits) 255 { 256 return (__m128i)__builtin_ia32_pslldi128((__v4si)__v, (int)__bits); 257 } 258 259 _INTRINSATTR 260 static __inline __m128i 261 _mm_slli_epi64(__m128i __v, uint8_t __bits) 262 { 263 return (__m128i)__builtin_ia32_psllqi128((__v2di)__v, (int)__bits); 264 } 265 266 #if defined(__GNUC__) && !defined(__clang__) 267 #define _mm_slli_si128(v,bytes) \ 268 (__m128i)__builtin_ia32_pslldqi128((__v2di)(__m128i)(v), \ 269 8*(int)(bytes)) 270 #elif defined(__clang__) 271 #define _mm_slli_si128(v,bytes) \ 272 (__m128i)__builtin_ia32_pslldqi128_byteshift((__v2di)(__m128i)(v), \ 273 (int)(bytes)) 274 #endif 275 276 _INTRINSATTR 277 static __inline __m128i 278 _mm_srli_epi32(__m128i __v, uint8_t __bits) 279 { 280 return (__m128i)__builtin_ia32_psrldi128((__v4si)__v, (int)__bits); 281 } 282 283 _INTRINSATTR 284 static __inline __m128i 285 _mm_srli_epi64(__m128i __v, uint8_t __bits) 286 { 287 return (__m128i)__builtin_ia32_psrlqi128((__v2di)__v, (int)__bits); 288 } 289 290 #if defined(__GNUC__) && !defined(__clang__) 291 #define _mm_srli_si128(v,bytes) \ 292 (__m128i)__builtin_ia32_psrldqi128((__m128i)(v), 8*(int)(bytes)) 293 #elif defined(__clang__) 294 #define _mm_srli_si128(v,bytes) \ 295 (__m128i)__builtin_ia32_psrldqi128_byteshift((__v2di)(__m128i)(v), \ 296 (int)(bytes)); 297 #endif 298 299 _INTRINSATTR 300 static __inline void 301 _mm_storeu_si128(__m128i_u *__p, __m128i __v) 302 { 303 #if defined(__GNUC__) && !defined(__clang__) 304 *__p = __v; 305 #else 306 ((struct { __m128i_u __v; } _PACKALIAS *)__p)->__v = __v; 307 #endif 308 } 309 310 _INTRINSATTR 311 static __inline void 312 _mm_storeu_si32(void *__p, __m128i __v) 313 { 314 #if defined(__GNUC__) && !defined(__clang__) 315 *(__m32_u *)__p = (__m32)((__v4si)__v)[0]; 316 #else 317 ((struct { int32_t __v; } _PACKALIAS *)__p)->__v = ((__v4si)__v)[0]; 318 #endif 319 } 320 321 _INTRINSATTR 322 static __inline void 323 _mm_storeu_si64(void *__p, __m128i __v) 324 { 325 #if defined(__GNUC__) && !defined(__clang__) 326 *(__m64_u *)__p = (__m64)((__v2di)__v)[0]; 327 #else 328 ((struct { int64_t __v; } _PACKALIAS *)__p)->__v = ((__v2di)__v)[0]; 329 #endif 330 } 331 332 _INTRINSATTR 333 static __inline void 334 _mm_store_si128(__m128i *__p, __m128i __v) 335 { 336 *__p = __v; 337 } 338 339 _INTRINSATTR 340 static __inline __m128i 341 _mm_sub_epi64(__m128i __x, __m128i __y) 342 { 343 return (__m128i)((__v2du)__x - (__v2du)__y); 344 } 345 346 _INTRINSATTR 347 static __inline __m128i 348 _mm_unpackhi_epi32(__m128i __lo, __m128i __hi) 349 { 350 #if defined(__GNUC__) && !defined(__clang__) 351 return (__m128i)__builtin_ia32_punpckhdq128((__v4si)__lo, 352 (__v4si)__hi); 353 #elif defined(__clang__) 354 return (__m128i)__builtin_shufflevector((__v4si)__lo, (__v4si)__hi, 355 2,6,3,7); 356 #endif 357 } 358 359 _INTRINSATTR 360 static __inline __m128i 361 _mm_unpacklo_epi32(__m128i __lo, __m128i __hi) 362 { 363 #if defined(__GNUC__) && !defined(__clang__) 364 return (__m128i)__builtin_ia32_punpckldq128((__v4si)__lo, 365 (__v4si)__hi); 366 #elif defined(__clang__) 367 return (__m128i)__builtin_shufflevector((__v4si)__lo, (__v4si)__hi, 368 0,4,1,5); 369 #endif 370 } 371 372 _INTRINSATTR 373 static __inline __m128i 374 _mm_unpacklo_epi64(__m128i __lo, __m128i __hi) 375 { 376 #if defined(__GNUC__) && !defined(__clang__) 377 return (__m128i)__builtin_ia32_punpcklqdq128((__v2di)__lo, 378 (__v2di)__hi); 379 #elif defined(__clang__) 380 return (__m128i)__builtin_shufflevector((__v2di)__lo, (__v2di)__hi, 381 0,2); 382 #endif 383 } 384 385 #endif /* _SYS_CRYPTO_ARCH_X86_IMMINTRIN_H */ 386
Indexes created Sat Sep 20 22:09:52 GMT 2025