1169689Skan/* Copyright (C) 2002, 2003, 2004 Free Software Foundation, Inc. 2132718Skan 3132718Skan This file is part of GCC. 4132718Skan 5132718Skan GCC is free software; you can redistribute it and/or modify it 6132718Skan under the terms of the GNU General Public License as published 7132718Skan by the Free Software Foundation; either version 2, or (at your 8132718Skan option) any later version. 9132718Skan 10132718Skan GCC is distributed in the hope that it will be useful, but WITHOUT 11132718Skan ANY WARRANTY; without even the implied warranty of MERCHANTABILITY 12132718Skan or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public 13132718Skan License for more details. 14132718Skan 15132718Skan You should have received a copy of the GNU General Public License 16132718Skan along with GCC; see the file COPYING. If not, write to the Free 17169689Skan Software Foundation, 51 Franklin Street, Fifth Floor, Boston, MA 18169689Skan 02110-1301, USA. */ 19132718Skan 20132718Skan/* As a special exception, if you include this header file into source 21132718Skan files compiled by GCC, this header file does not by itself cause 22132718Skan the resulting executable to be covered by the GNU General Public 23132718Skan License. This exception does not however invalidate any other 24132718Skan reasons why the executable file might be covered by the GNU General 25132718Skan Public License. */ 26132718Skan 27132718Skan#ifndef _MMINTRIN_H_INCLUDED 28132718Skan#define _MMINTRIN_H_INCLUDED 29132718Skan 30132718Skan/* The data type intended for user use. */ 31132718Skantypedef unsigned long long __m64, __int64; 32132718Skan 33132718Skan/* Internal data types for implementing the intrinsics. */ 34169689Skantypedef int __v2si __attribute__ ((vector_size (8))); 35169689Skantypedef short __v4hi __attribute__ ((vector_size (8))); 36169689Skantypedef char __v8qi __attribute__ ((vector_size (8))); 37132718Skan 38132718Skan/* "Convert" __m64 and __int64 into each other. */ 39132718Skanstatic __inline __m64 40132718Skan_mm_cvtsi64_m64 (__int64 __i) 41132718Skan{ 42132718Skan return __i; 43132718Skan} 44132718Skan 45132718Skanstatic __inline __int64 46132718Skan_mm_cvtm64_si64 (__m64 __i) 47132718Skan{ 48132718Skan return __i; 49132718Skan} 50132718Skan 51132718Skanstatic __inline int 52132718Skan_mm_cvtsi64_si32 (__int64 __i) 53132718Skan{ 54132718Skan return __i; 55132718Skan} 56132718Skan 57132718Skanstatic __inline __int64 58132718Skan_mm_cvtsi32_si64 (int __i) 59132718Skan{ 60132718Skan return __i; 61132718Skan} 62132718Skan 63132718Skan/* Pack the four 16-bit values from M1 into the lower four 8-bit values of 64132718Skan the result, and the four 16-bit values from M2 into the upper four 8-bit 65132718Skan values of the result, all with signed saturation. */ 66132718Skanstatic __inline __m64 67132718Skan_mm_packs_pi16 (__m64 __m1, __m64 __m2) 68132718Skan{ 69132718Skan return (__m64) __builtin_arm_wpackhss ((__v4hi)__m1, (__v4hi)__m2); 70132718Skan} 71132718Skan 72132718Skan/* Pack the two 32-bit values from M1 in to the lower two 16-bit values of 73132718Skan the result, and the two 32-bit values from M2 into the upper two 16-bit 74132718Skan values of the result, all with signed saturation. */ 75132718Skanstatic __inline __m64 76132718Skan_mm_packs_pi32 (__m64 __m1, __m64 __m2) 77132718Skan{ 78132718Skan return (__m64) __builtin_arm_wpackwss ((__v2si)__m1, (__v2si)__m2); 79132718Skan} 80132718Skan 81132718Skan/* Copy the 64-bit value from M1 into the lower 32-bits of the result, and 82132718Skan the 64-bit value from M2 into the upper 32-bits of the result, all with 83132718Skan signed saturation for values that do not fit exactly into 32-bits. */ 84132718Skanstatic __inline __m64 85132718Skan_mm_packs_pi64 (__m64 __m1, __m64 __m2) 86132718Skan{ 87132718Skan return (__m64) __builtin_arm_wpackdss ((long long)__m1, (long long)__m2); 88132718Skan} 89132718Skan 90132718Skan/* Pack the four 16-bit values from M1 into the lower four 8-bit values of 91132718Skan the result, and the four 16-bit values from M2 into the upper four 8-bit 92132718Skan values of the result, all with unsigned saturation. */ 93132718Skanstatic __inline __m64 94132718Skan_mm_packs_pu16 (__m64 __m1, __m64 __m2) 95132718Skan{ 96132718Skan return (__m64) __builtin_arm_wpackhus ((__v4hi)__m1, (__v4hi)__m2); 97132718Skan} 98132718Skan 99132718Skan/* Pack the two 32-bit values from M1 into the lower two 16-bit values of 100132718Skan the result, and the two 32-bit values from M2 into the upper two 16-bit 101132718Skan values of the result, all with unsigned saturation. */ 102132718Skanstatic __inline __m64 103132718Skan_mm_packs_pu32 (__m64 __m1, __m64 __m2) 104132718Skan{ 105132718Skan return (__m64) __builtin_arm_wpackwus ((__v2si)__m1, (__v2si)__m2); 106132718Skan} 107132718Skan 108132718Skan/* Copy the 64-bit value from M1 into the lower 32-bits of the result, and 109132718Skan the 64-bit value from M2 into the upper 32-bits of the result, all with 110132718Skan unsigned saturation for values that do not fit exactly into 32-bits. */ 111132718Skanstatic __inline __m64 112132718Skan_mm_packs_pu64 (__m64 __m1, __m64 __m2) 113132718Skan{ 114132718Skan return (__m64) __builtin_arm_wpackdus ((long long)__m1, (long long)__m2); 115132718Skan} 116132718Skan 117132718Skan/* Interleave the four 8-bit values from the high half of M1 with the four 118132718Skan 8-bit values from the high half of M2. */ 119132718Skanstatic __inline __m64 120132718Skan_mm_unpackhi_pi8 (__m64 __m1, __m64 __m2) 121132718Skan{ 122132718Skan return (__m64) __builtin_arm_wunpckihb ((__v8qi)__m1, (__v8qi)__m2); 123132718Skan} 124132718Skan 125132718Skan/* Interleave the two 16-bit values from the high half of M1 with the two 126132718Skan 16-bit values from the high half of M2. */ 127132718Skanstatic __inline __m64 128132718Skan_mm_unpackhi_pi16 (__m64 __m1, __m64 __m2) 129132718Skan{ 130132718Skan return (__m64) __builtin_arm_wunpckihh ((__v4hi)__m1, (__v4hi)__m2); 131132718Skan} 132132718Skan 133132718Skan/* Interleave the 32-bit value from the high half of M1 with the 32-bit 134132718Skan value from the high half of M2. */ 135132718Skanstatic __inline __m64 136132718Skan_mm_unpackhi_pi32 (__m64 __m1, __m64 __m2) 137132718Skan{ 138132718Skan return (__m64) __builtin_arm_wunpckihw ((__v2si)__m1, (__v2si)__m2); 139132718Skan} 140132718Skan 141132718Skan/* Interleave the four 8-bit values from the low half of M1 with the four 142132718Skan 8-bit values from the low half of M2. */ 143132718Skanstatic __inline __m64 144132718Skan_mm_unpacklo_pi8 (__m64 __m1, __m64 __m2) 145132718Skan{ 146132718Skan return (__m64) __builtin_arm_wunpckilb ((__v8qi)__m1, (__v8qi)__m2); 147132718Skan} 148132718Skan 149132718Skan/* Interleave the two 16-bit values from the low half of M1 with the two 150132718Skan 16-bit values from the low half of M2. */ 151132718Skanstatic __inline __m64 152132718Skan_mm_unpacklo_pi16 (__m64 __m1, __m64 __m2) 153132718Skan{ 154132718Skan return (__m64) __builtin_arm_wunpckilh ((__v4hi)__m1, (__v4hi)__m2); 155132718Skan} 156132718Skan 157132718Skan/* Interleave the 32-bit value from the low half of M1 with the 32-bit 158132718Skan value from the low half of M2. */ 159132718Skanstatic __inline __m64 160132718Skan_mm_unpacklo_pi32 (__m64 __m1, __m64 __m2) 161132718Skan{ 162132718Skan return (__m64) __builtin_arm_wunpckilw ((__v2si)__m1, (__v2si)__m2); 163132718Skan} 164132718Skan 165132718Skan/* Take the four 8-bit values from the low half of M1, sign extend them, 166132718Skan and return the result as a vector of four 16-bit quantities. */ 167132718Skanstatic __inline __m64 168132718Skan_mm_unpackel_pi8 (__m64 __m1) 169132718Skan{ 170132718Skan return (__m64) __builtin_arm_wunpckelsb ((__v8qi)__m1); 171132718Skan} 172132718Skan 173132718Skan/* Take the two 16-bit values from the low half of M1, sign extend them, 174132718Skan and return the result as a vector of two 32-bit quantities. */ 175132718Skanstatic __inline __m64 176132718Skan_mm_unpackel_pi16 (__m64 __m1) 177132718Skan{ 178132718Skan return (__m64) __builtin_arm_wunpckelsh ((__v4hi)__m1); 179132718Skan} 180132718Skan 181132718Skan/* Take the 32-bit value from the low half of M1, and return it sign extended 182132718Skan to 64 bits. */ 183132718Skanstatic __inline __m64 184132718Skan_mm_unpackel_pi32 (__m64 __m1) 185132718Skan{ 186132718Skan return (__m64) __builtin_arm_wunpckelsw ((__v2si)__m1); 187132718Skan} 188132718Skan 189132718Skan/* Take the four 8-bit values from the high half of M1, sign extend them, 190132718Skan and return the result as a vector of four 16-bit quantities. */ 191132718Skanstatic __inline __m64 192132718Skan_mm_unpackeh_pi8 (__m64 __m1) 193132718Skan{ 194132718Skan return (__m64) __builtin_arm_wunpckehsb ((__v8qi)__m1); 195132718Skan} 196132718Skan 197132718Skan/* Take the two 16-bit values from the high half of M1, sign extend them, 198132718Skan and return the result as a vector of two 32-bit quantities. */ 199132718Skanstatic __inline __m64 200132718Skan_mm_unpackeh_pi16 (__m64 __m1) 201132718Skan{ 202132718Skan return (__m64) __builtin_arm_wunpckehsh ((__v4hi)__m1); 203132718Skan} 204132718Skan 205132718Skan/* Take the 32-bit value from the high half of M1, and return it sign extended 206132718Skan to 64 bits. */ 207132718Skanstatic __inline __m64 208132718Skan_mm_unpackeh_pi32 (__m64 __m1) 209132718Skan{ 210132718Skan return (__m64) __builtin_arm_wunpckehsw ((__v2si)__m1); 211132718Skan} 212132718Skan 213132718Skan/* Take the four 8-bit values from the low half of M1, zero extend them, 214132718Skan and return the result as a vector of four 16-bit quantities. */ 215132718Skanstatic __inline __m64 216132718Skan_mm_unpackel_pu8 (__m64 __m1) 217132718Skan{ 218132718Skan return (__m64) __builtin_arm_wunpckelub ((__v8qi)__m1); 219132718Skan} 220132718Skan 221132718Skan/* Take the two 16-bit values from the low half of M1, zero extend them, 222132718Skan and return the result as a vector of two 32-bit quantities. */ 223132718Skanstatic __inline __m64 224132718Skan_mm_unpackel_pu16 (__m64 __m1) 225132718Skan{ 226132718Skan return (__m64) __builtin_arm_wunpckeluh ((__v4hi)__m1); 227132718Skan} 228132718Skan 229132718Skan/* Take the 32-bit value from the low half of M1, and return it zero extended 230132718Skan to 64 bits. */ 231132718Skanstatic __inline __m64 232132718Skan_mm_unpackel_pu32 (__m64 __m1) 233132718Skan{ 234132718Skan return (__m64) __builtin_arm_wunpckeluw ((__v2si)__m1); 235132718Skan} 236132718Skan 237132718Skan/* Take the four 8-bit values from the high half of M1, zero extend them, 238132718Skan and return the result as a vector of four 16-bit quantities. */ 239132718Skanstatic __inline __m64 240132718Skan_mm_unpackeh_pu8 (__m64 __m1) 241132718Skan{ 242132718Skan return (__m64) __builtin_arm_wunpckehub ((__v8qi)__m1); 243132718Skan} 244132718Skan 245132718Skan/* Take the two 16-bit values from the high half of M1, zero extend them, 246132718Skan and return the result as a vector of two 32-bit quantities. */ 247132718Skanstatic __inline __m64 248132718Skan_mm_unpackeh_pu16 (__m64 __m1) 249132718Skan{ 250132718Skan return (__m64) __builtin_arm_wunpckehuh ((__v4hi)__m1); 251132718Skan} 252132718Skan 253132718Skan/* Take the 32-bit value from the high half of M1, and return it zero extended 254132718Skan to 64 bits. */ 255132718Skanstatic __inline __m64 256132718Skan_mm_unpackeh_pu32 (__m64 __m1) 257132718Skan{ 258132718Skan return (__m64) __builtin_arm_wunpckehuw ((__v2si)__m1); 259132718Skan} 260132718Skan 261132718Skan/* Add the 8-bit values in M1 to the 8-bit values in M2. */ 262132718Skanstatic __inline __m64 263132718Skan_mm_add_pi8 (__m64 __m1, __m64 __m2) 264132718Skan{ 265132718Skan return (__m64) __builtin_arm_waddb ((__v8qi)__m1, (__v8qi)__m2); 266132718Skan} 267132718Skan 268132718Skan/* Add the 16-bit values in M1 to the 16-bit values in M2. */ 269132718Skanstatic __inline __m64 270132718Skan_mm_add_pi16 (__m64 __m1, __m64 __m2) 271132718Skan{ 272132718Skan return (__m64) __builtin_arm_waddh ((__v4hi)__m1, (__v4hi)__m2); 273132718Skan} 274132718Skan 275132718Skan/* Add the 32-bit values in M1 to the 32-bit values in M2. */ 276132718Skanstatic __inline __m64 277132718Skan_mm_add_pi32 (__m64 __m1, __m64 __m2) 278132718Skan{ 279132718Skan return (__m64) __builtin_arm_waddw ((__v2si)__m1, (__v2si)__m2); 280132718Skan} 281132718Skan 282132718Skan/* Add the 8-bit values in M1 to the 8-bit values in M2 using signed 283132718Skan saturated arithmetic. */ 284132718Skanstatic __inline __m64 285132718Skan_mm_adds_pi8 (__m64 __m1, __m64 __m2) 286132718Skan{ 287132718Skan return (__m64) __builtin_arm_waddbss ((__v8qi)__m1, (__v8qi)__m2); 288132718Skan} 289132718Skan 290132718Skan/* Add the 16-bit values in M1 to the 16-bit values in M2 using signed 291132718Skan saturated arithmetic. */ 292132718Skanstatic __inline __m64 293132718Skan_mm_adds_pi16 (__m64 __m1, __m64 __m2) 294132718Skan{ 295132718Skan return (__m64) __builtin_arm_waddhss ((__v4hi)__m1, (__v4hi)__m2); 296132718Skan} 297132718Skan 298132718Skan/* Add the 32-bit values in M1 to the 32-bit values in M2 using signed 299132718Skan saturated arithmetic. */ 300132718Skanstatic __inline __m64 301132718Skan_mm_adds_pi32 (__m64 __m1, __m64 __m2) 302132718Skan{ 303132718Skan return (__m64) __builtin_arm_waddwss ((__v2si)__m1, (__v2si)__m2); 304132718Skan} 305132718Skan 306132718Skan/* Add the 8-bit values in M1 to the 8-bit values in M2 using unsigned 307132718Skan saturated arithmetic. */ 308132718Skanstatic __inline __m64 309132718Skan_mm_adds_pu8 (__m64 __m1, __m64 __m2) 310132718Skan{ 311132718Skan return (__m64) __builtin_arm_waddbus ((__v8qi)__m1, (__v8qi)__m2); 312132718Skan} 313132718Skan 314132718Skan/* Add the 16-bit values in M1 to the 16-bit values in M2 using unsigned 315132718Skan saturated arithmetic. */ 316132718Skanstatic __inline __m64 317132718Skan_mm_adds_pu16 (__m64 __m1, __m64 __m2) 318132718Skan{ 319132718Skan return (__m64) __builtin_arm_waddhus ((__v4hi)__m1, (__v4hi)__m2); 320132718Skan} 321132718Skan 322132718Skan/* Add the 32-bit values in M1 to the 32-bit values in M2 using unsigned 323132718Skan saturated arithmetic. */ 324132718Skanstatic __inline __m64 325132718Skan_mm_adds_pu32 (__m64 __m1, __m64 __m2) 326132718Skan{ 327132718Skan return (__m64) __builtin_arm_waddwus ((__v2si)__m1, (__v2si)__m2); 328132718Skan} 329132718Skan 330132718Skan/* Subtract the 8-bit values in M2 from the 8-bit values in M1. */ 331132718Skanstatic __inline __m64 332132718Skan_mm_sub_pi8 (__m64 __m1, __m64 __m2) 333132718Skan{ 334132718Skan return (__m64) __builtin_arm_wsubb ((__v8qi)__m1, (__v8qi)__m2); 335132718Skan} 336132718Skan 337132718Skan/* Subtract the 16-bit values in M2 from the 16-bit values in M1. */ 338132718Skanstatic __inline __m64 339132718Skan_mm_sub_pi16 (__m64 __m1, __m64 __m2) 340132718Skan{ 341132718Skan return (__m64) __builtin_arm_wsubh ((__v4hi)__m1, (__v4hi)__m2); 342132718Skan} 343132718Skan 344132718Skan/* Subtract the 32-bit values in M2 from the 32-bit values in M1. */ 345132718Skanstatic __inline __m64 346132718Skan_mm_sub_pi32 (__m64 __m1, __m64 __m2) 347132718Skan{ 348132718Skan return (__m64) __builtin_arm_wsubw ((__v2si)__m1, (__v2si)__m2); 349132718Skan} 350132718Skan 351132718Skan/* Subtract the 8-bit values in M2 from the 8-bit values in M1 using signed 352132718Skan saturating arithmetic. */ 353132718Skanstatic __inline __m64 354132718Skan_mm_subs_pi8 (__m64 __m1, __m64 __m2) 355132718Skan{ 356132718Skan return (__m64) __builtin_arm_wsubbss ((__v8qi)__m1, (__v8qi)__m2); 357132718Skan} 358132718Skan 359132718Skan/* Subtract the 16-bit values in M2 from the 16-bit values in M1 using 360132718Skan signed saturating arithmetic. */ 361132718Skanstatic __inline __m64 362132718Skan_mm_subs_pi16 (__m64 __m1, __m64 __m2) 363132718Skan{ 364132718Skan return (__m64) __builtin_arm_wsubhss ((__v4hi)__m1, (__v4hi)__m2); 365132718Skan} 366132718Skan 367132718Skan/* Subtract the 32-bit values in M2 from the 32-bit values in M1 using 368132718Skan signed saturating arithmetic. */ 369132718Skanstatic __inline __m64 370132718Skan_mm_subs_pi32 (__m64 __m1, __m64 __m2) 371132718Skan{ 372132718Skan return (__m64) __builtin_arm_wsubwss ((__v2si)__m1, (__v2si)__m2); 373132718Skan} 374132718Skan 375132718Skan/* Subtract the 8-bit values in M2 from the 8-bit values in M1 using 376132718Skan unsigned saturating arithmetic. */ 377132718Skanstatic __inline __m64 378132718Skan_mm_subs_pu8 (__m64 __m1, __m64 __m2) 379132718Skan{ 380132718Skan return (__m64) __builtin_arm_wsubbus ((__v8qi)__m1, (__v8qi)__m2); 381132718Skan} 382132718Skan 383132718Skan/* Subtract the 16-bit values in M2 from the 16-bit values in M1 using 384132718Skan unsigned saturating arithmetic. */ 385132718Skanstatic __inline __m64 386132718Skan_mm_subs_pu16 (__m64 __m1, __m64 __m2) 387132718Skan{ 388132718Skan return (__m64) __builtin_arm_wsubhus ((__v4hi)__m1, (__v4hi)__m2); 389132718Skan} 390132718Skan 391132718Skan/* Subtract the 32-bit values in M2 from the 32-bit values in M1 using 392132718Skan unsigned saturating arithmetic. */ 393132718Skanstatic __inline __m64 394132718Skan_mm_subs_pu32 (__m64 __m1, __m64 __m2) 395132718Skan{ 396132718Skan return (__m64) __builtin_arm_wsubwus ((__v2si)__m1, (__v2si)__m2); 397132718Skan} 398132718Skan 399132718Skan/* Multiply four 16-bit values in M1 by four 16-bit values in M2 producing 400132718Skan four 32-bit intermediate results, which are then summed by pairs to 401132718Skan produce two 32-bit results. */ 402132718Skanstatic __inline __m64 403132718Skan_mm_madd_pi16 (__m64 __m1, __m64 __m2) 404132718Skan{ 405132718Skan return (__m64) __builtin_arm_wmadds ((__v4hi)__m1, (__v4hi)__m2); 406132718Skan} 407132718Skan 408132718Skan/* Multiply four 16-bit values in M1 by four 16-bit values in M2 producing 409132718Skan four 32-bit intermediate results, which are then summed by pairs to 410132718Skan produce two 32-bit results. */ 411132718Skanstatic __inline __m64 412132718Skan_mm_madd_pu16 (__m64 __m1, __m64 __m2) 413132718Skan{ 414132718Skan return (__m64) __builtin_arm_wmaddu ((__v4hi)__m1, (__v4hi)__m2); 415132718Skan} 416132718Skan 417132718Skan/* Multiply four signed 16-bit values in M1 by four signed 16-bit values in 418132718Skan M2 and produce the high 16 bits of the 32-bit results. */ 419132718Skanstatic __inline __m64 420132718Skan_mm_mulhi_pi16 (__m64 __m1, __m64 __m2) 421132718Skan{ 422169689Skan return (__m64) __builtin_arm_wmulsm ((__v4hi)__m1, (__v4hi)__m2); 423132718Skan} 424132718Skan 425132718Skan/* Multiply four signed 16-bit values in M1 by four signed 16-bit values in 426132718Skan M2 and produce the high 16 bits of the 32-bit results. */ 427132718Skanstatic __inline __m64 428132718Skan_mm_mulhi_pu16 (__m64 __m1, __m64 __m2) 429132718Skan{ 430169689Skan return (__m64) __builtin_arm_wmulum ((__v4hi)__m1, (__v4hi)__m2); 431132718Skan} 432132718Skan 433132718Skan/* Multiply four 16-bit values in M1 by four 16-bit values in M2 and produce 434132718Skan the low 16 bits of the results. */ 435132718Skanstatic __inline __m64 436132718Skan_mm_mullo_pi16 (__m64 __m1, __m64 __m2) 437132718Skan{ 438132718Skan return (__m64) __builtin_arm_wmulul ((__v4hi)__m1, (__v4hi)__m2); 439132718Skan} 440132718Skan 441132718Skan/* Shift four 16-bit values in M left by COUNT. */ 442132718Skanstatic __inline __m64 443132718Skan_mm_sll_pi16 (__m64 __m, __m64 __count) 444132718Skan{ 445132718Skan return (__m64) __builtin_arm_wsllh ((__v4hi)__m, __count); 446132718Skan} 447132718Skan 448132718Skanstatic __inline __m64 449132718Skan_mm_slli_pi16 (__m64 __m, int __count) 450132718Skan{ 451132718Skan return (__m64) __builtin_arm_wsllhi ((__v4hi)__m, __count); 452132718Skan} 453132718Skan 454132718Skan/* Shift two 32-bit values in M left by COUNT. */ 455132718Skanstatic __inline __m64 456132718Skan_mm_sll_pi32 (__m64 __m, __m64 __count) 457132718Skan{ 458132718Skan return (__m64) __builtin_arm_wsllw ((__v2si)__m, __count); 459132718Skan} 460132718Skan 461132718Skanstatic __inline __m64 462132718Skan_mm_slli_pi32 (__m64 __m, int __count) 463132718Skan{ 464132718Skan return (__m64) __builtin_arm_wsllwi ((__v2si)__m, __count); 465132718Skan} 466132718Skan 467132718Skan/* Shift the 64-bit value in M left by COUNT. */ 468132718Skanstatic __inline __m64 469132718Skan_mm_sll_si64 (__m64 __m, __m64 __count) 470132718Skan{ 471132718Skan return (__m64) __builtin_arm_wslld (__m, __count); 472132718Skan} 473132718Skan 474132718Skanstatic __inline __m64 475132718Skan_mm_slli_si64 (__m64 __m, int __count) 476132718Skan{ 477132718Skan return (__m64) __builtin_arm_wslldi (__m, __count); 478132718Skan} 479132718Skan 480132718Skan/* Shift four 16-bit values in M right by COUNT; shift in the sign bit. */ 481132718Skanstatic __inline __m64 482132718Skan_mm_sra_pi16 (__m64 __m, __m64 __count) 483132718Skan{ 484132718Skan return (__m64) __builtin_arm_wsrah ((__v4hi)__m, __count); 485132718Skan} 486132718Skan 487132718Skanstatic __inline __m64 488132718Skan_mm_srai_pi16 (__m64 __m, int __count) 489132718Skan{ 490132718Skan return (__m64) __builtin_arm_wsrahi ((__v4hi)__m, __count); 491132718Skan} 492132718Skan 493132718Skan/* Shift two 32-bit values in M right by COUNT; shift in the sign bit. */ 494132718Skanstatic __inline __m64 495132718Skan_mm_sra_pi32 (__m64 __m, __m64 __count) 496132718Skan{ 497132718Skan return (__m64) __builtin_arm_wsraw ((__v2si)__m, __count); 498132718Skan} 499132718Skan 500132718Skanstatic __inline __m64 501132718Skan_mm_srai_pi32 (__m64 __m, int __count) 502132718Skan{ 503132718Skan return (__m64) __builtin_arm_wsrawi ((__v2si)__m, __count); 504132718Skan} 505132718Skan 506132718Skan/* Shift the 64-bit value in M right by COUNT; shift in the sign bit. */ 507132718Skanstatic __inline __m64 508132718Skan_mm_sra_si64 (__m64 __m, __m64 __count) 509132718Skan{ 510132718Skan return (__m64) __builtin_arm_wsrad (__m, __count); 511132718Skan} 512132718Skan 513132718Skanstatic __inline __m64 514132718Skan_mm_srai_si64 (__m64 __m, int __count) 515132718Skan{ 516132718Skan return (__m64) __builtin_arm_wsradi (__m, __count); 517132718Skan} 518132718Skan 519132718Skan/* Shift four 16-bit values in M right by COUNT; shift in zeros. */ 520132718Skanstatic __inline __m64 521132718Skan_mm_srl_pi16 (__m64 __m, __m64 __count) 522132718Skan{ 523132718Skan return (__m64) __builtin_arm_wsrlh ((__v4hi)__m, __count); 524132718Skan} 525132718Skan 526132718Skanstatic __inline __m64 527132718Skan_mm_srli_pi16 (__m64 __m, int __count) 528132718Skan{ 529132718Skan return (__m64) __builtin_arm_wsrlhi ((__v4hi)__m, __count); 530132718Skan} 531132718Skan 532132718Skan/* Shift two 32-bit values in M right by COUNT; shift in zeros. */ 533132718Skanstatic __inline __m64 534132718Skan_mm_srl_pi32 (__m64 __m, __m64 __count) 535132718Skan{ 536132718Skan return (__m64) __builtin_arm_wsrlw ((__v2si)__m, __count); 537132718Skan} 538132718Skan 539132718Skanstatic __inline __m64 540132718Skan_mm_srli_pi32 (__m64 __m, int __count) 541132718Skan{ 542132718Skan return (__m64) __builtin_arm_wsrlwi ((__v2si)__m, __count); 543132718Skan} 544132718Skan 545132718Skan/* Shift the 64-bit value in M left by COUNT; shift in zeros. */ 546132718Skanstatic __inline __m64 547132718Skan_mm_srl_si64 (__m64 __m, __m64 __count) 548132718Skan{ 549132718Skan return (__m64) __builtin_arm_wsrld (__m, __count); 550132718Skan} 551132718Skan 552132718Skanstatic __inline __m64 553132718Skan_mm_srli_si64 (__m64 __m, int __count) 554132718Skan{ 555132718Skan return (__m64) __builtin_arm_wsrldi (__m, __count); 556132718Skan} 557132718Skan 558132718Skan/* Rotate four 16-bit values in M right by COUNT. */ 559132718Skanstatic __inline __m64 560132718Skan_mm_ror_pi16 (__m64 __m, __m64 __count) 561132718Skan{ 562132718Skan return (__m64) __builtin_arm_wrorh ((__v4hi)__m, __count); 563132718Skan} 564132718Skan 565132718Skanstatic __inline __m64 566132718Skan_mm_rori_pi16 (__m64 __m, int __count) 567132718Skan{ 568132718Skan return (__m64) __builtin_arm_wrorhi ((__v4hi)__m, __count); 569132718Skan} 570132718Skan 571132718Skan/* Rotate two 32-bit values in M right by COUNT. */ 572132718Skanstatic __inline __m64 573132718Skan_mm_ror_pi32 (__m64 __m, __m64 __count) 574132718Skan{ 575132718Skan return (__m64) __builtin_arm_wrorw ((__v2si)__m, __count); 576132718Skan} 577132718Skan 578132718Skanstatic __inline __m64 579132718Skan_mm_rori_pi32 (__m64 __m, int __count) 580132718Skan{ 581132718Skan return (__m64) __builtin_arm_wrorwi ((__v2si)__m, __count); 582132718Skan} 583132718Skan 584132718Skan/* Rotate two 64-bit values in M right by COUNT. */ 585132718Skanstatic __inline __m64 586132718Skan_mm_ror_si64 (__m64 __m, __m64 __count) 587132718Skan{ 588132718Skan return (__m64) __builtin_arm_wrord (__m, __count); 589132718Skan} 590132718Skan 591132718Skanstatic __inline __m64 592132718Skan_mm_rori_si64 (__m64 __m, int __count) 593132718Skan{ 594132718Skan return (__m64) __builtin_arm_wrordi (__m, __count); 595132718Skan} 596132718Skan 597132718Skan/* Bit-wise AND the 64-bit values in M1 and M2. */ 598132718Skanstatic __inline __m64 599132718Skan_mm_and_si64 (__m64 __m1, __m64 __m2) 600132718Skan{ 601132718Skan return __builtin_arm_wand (__m1, __m2); 602132718Skan} 603132718Skan 604132718Skan/* Bit-wise complement the 64-bit value in M1 and bit-wise AND it with the 605132718Skan 64-bit value in M2. */ 606132718Skanstatic __inline __m64 607132718Skan_mm_andnot_si64 (__m64 __m1, __m64 __m2) 608132718Skan{ 609132718Skan return __builtin_arm_wandn (__m1, __m2); 610132718Skan} 611132718Skan 612132718Skan/* Bit-wise inclusive OR the 64-bit values in M1 and M2. */ 613132718Skanstatic __inline __m64 614132718Skan_mm_or_si64 (__m64 __m1, __m64 __m2) 615132718Skan{ 616132718Skan return __builtin_arm_wor (__m1, __m2); 617132718Skan} 618132718Skan 619132718Skan/* Bit-wise exclusive OR the 64-bit values in M1 and M2. */ 620132718Skanstatic __inline __m64 621132718Skan_mm_xor_si64 (__m64 __m1, __m64 __m2) 622132718Skan{ 623132718Skan return __builtin_arm_wxor (__m1, __m2); 624132718Skan} 625132718Skan 626132718Skan/* Compare eight 8-bit values. The result of the comparison is 0xFF if the 627132718Skan test is true and zero if false. */ 628132718Skanstatic __inline __m64 629132718Skan_mm_cmpeq_pi8 (__m64 __m1, __m64 __m2) 630132718Skan{ 631132718Skan return (__m64) __builtin_arm_wcmpeqb ((__v8qi)__m1, (__v8qi)__m2); 632132718Skan} 633132718Skan 634132718Skanstatic __inline __m64 635132718Skan_mm_cmpgt_pi8 (__m64 __m1, __m64 __m2) 636132718Skan{ 637132718Skan return (__m64) __builtin_arm_wcmpgtsb ((__v8qi)__m1, (__v8qi)__m2); 638132718Skan} 639132718Skan 640132718Skanstatic __inline __m64 641132718Skan_mm_cmpgt_pu8 (__m64 __m1, __m64 __m2) 642132718Skan{ 643132718Skan return (__m64) __builtin_arm_wcmpgtub ((__v8qi)__m1, (__v8qi)__m2); 644132718Skan} 645132718Skan 646132718Skan/* Compare four 16-bit values. The result of the comparison is 0xFFFF if 647132718Skan the test is true and zero if false. */ 648132718Skanstatic __inline __m64 649132718Skan_mm_cmpeq_pi16 (__m64 __m1, __m64 __m2) 650132718Skan{ 651132718Skan return (__m64) __builtin_arm_wcmpeqh ((__v4hi)__m1, (__v4hi)__m2); 652132718Skan} 653132718Skan 654132718Skanstatic __inline __m64 655132718Skan_mm_cmpgt_pi16 (__m64 __m1, __m64 __m2) 656132718Skan{ 657132718Skan return (__m64) __builtin_arm_wcmpgtsh ((__v4hi)__m1, (__v4hi)__m2); 658132718Skan} 659132718Skan 660132718Skanstatic __inline __m64 661132718Skan_mm_cmpgt_pu16 (__m64 __m1, __m64 __m2) 662132718Skan{ 663132718Skan return (__m64) __builtin_arm_wcmpgtuh ((__v4hi)__m1, (__v4hi)__m2); 664132718Skan} 665132718Skan 666132718Skan/* Compare two 32-bit values. The result of the comparison is 0xFFFFFFFF if 667132718Skan the test is true and zero if false. */ 668132718Skanstatic __inline __m64 669132718Skan_mm_cmpeq_pi32 (__m64 __m1, __m64 __m2) 670132718Skan{ 671132718Skan return (__m64) __builtin_arm_wcmpeqw ((__v2si)__m1, (__v2si)__m2); 672132718Skan} 673132718Skan 674132718Skanstatic __inline __m64 675132718Skan_mm_cmpgt_pi32 (__m64 __m1, __m64 __m2) 676132718Skan{ 677132718Skan return (__m64) __builtin_arm_wcmpgtsw ((__v2si)__m1, (__v2si)__m2); 678132718Skan} 679132718Skan 680132718Skanstatic __inline __m64 681132718Skan_mm_cmpgt_pu32 (__m64 __m1, __m64 __m2) 682132718Skan{ 683132718Skan return (__m64) __builtin_arm_wcmpgtuw ((__v2si)__m1, (__v2si)__m2); 684132718Skan} 685132718Skan 686132718Skan/* Element-wise multiplication of unsigned 16-bit values __B and __C, followed 687132718Skan by accumulate across all elements and __A. */ 688132718Skanstatic __inline __m64 689132718Skan_mm_mac_pu16 (__m64 __A, __m64 __B, __m64 __C) 690132718Skan{ 691132718Skan return __builtin_arm_wmacu (__A, (__v4hi)__B, (__v4hi)__C); 692132718Skan} 693132718Skan 694132718Skan/* Element-wise multiplication of signed 16-bit values __B and __C, followed 695132718Skan by accumulate across all elements and __A. */ 696132718Skanstatic __inline __m64 697132718Skan_mm_mac_pi16 (__m64 __A, __m64 __B, __m64 __C) 698132718Skan{ 699132718Skan return __builtin_arm_wmacs (__A, (__v4hi)__B, (__v4hi)__C); 700132718Skan} 701132718Skan 702132718Skan/* Element-wise multiplication of unsigned 16-bit values __B and __C, followed 703132718Skan by accumulate across all elements. */ 704132718Skanstatic __inline __m64 705132718Skan_mm_macz_pu16 (__m64 __A, __m64 __B) 706132718Skan{ 707132718Skan return __builtin_arm_wmacuz ((__v4hi)__A, (__v4hi)__B); 708132718Skan} 709132718Skan 710132718Skan/* Element-wise multiplication of signed 16-bit values __B and __C, followed 711132718Skan by accumulate across all elements. */ 712132718Skanstatic __inline __m64 713132718Skan_mm_macz_pi16 (__m64 __A, __m64 __B) 714132718Skan{ 715132718Skan return __builtin_arm_wmacsz ((__v4hi)__A, (__v4hi)__B); 716132718Skan} 717132718Skan 718132718Skan/* Accumulate across all unsigned 8-bit values in __A. */ 719132718Skanstatic __inline __m64 720132718Skan_mm_acc_pu8 (__m64 __A) 721132718Skan{ 722132718Skan return __builtin_arm_waccb ((__v8qi)__A); 723132718Skan} 724132718Skan 725132718Skan/* Accumulate across all unsigned 16-bit values in __A. */ 726132718Skanstatic __inline __m64 727132718Skan_mm_acc_pu16 (__m64 __A) 728132718Skan{ 729132718Skan return __builtin_arm_wacch ((__v4hi)__A); 730132718Skan} 731132718Skan 732132718Skan/* Accumulate across all unsigned 32-bit values in __A. */ 733132718Skanstatic __inline __m64 734132718Skan_mm_acc_pu32 (__m64 __A) 735132718Skan{ 736132718Skan return __builtin_arm_waccw ((__v2si)__A); 737132718Skan} 738132718Skan 739132718Skanstatic __inline __m64 740132718Skan_mm_mia_si64 (__m64 __A, int __B, int __C) 741132718Skan{ 742132718Skan return __builtin_arm_tmia (__A, __B, __C); 743132718Skan} 744132718Skan 745132718Skanstatic __inline __m64 746132718Skan_mm_miaph_si64 (__m64 __A, int __B, int __C) 747132718Skan{ 748132718Skan return __builtin_arm_tmiaph (__A, __B, __C); 749132718Skan} 750132718Skan 751132718Skanstatic __inline __m64 752132718Skan_mm_miabb_si64 (__m64 __A, int __B, int __C) 753132718Skan{ 754132718Skan return __builtin_arm_tmiabb (__A, __B, __C); 755132718Skan} 756132718Skan 757132718Skanstatic __inline __m64 758132718Skan_mm_miabt_si64 (__m64 __A, int __B, int __C) 759132718Skan{ 760132718Skan return __builtin_arm_tmiabt (__A, __B, __C); 761132718Skan} 762132718Skan 763132718Skanstatic __inline __m64 764132718Skan_mm_miatb_si64 (__m64 __A, int __B, int __C) 765132718Skan{ 766132718Skan return __builtin_arm_tmiatb (__A, __B, __C); 767132718Skan} 768132718Skan 769132718Skanstatic __inline __m64 770132718Skan_mm_miatt_si64 (__m64 __A, int __B, int __C) 771132718Skan{ 772132718Skan return __builtin_arm_tmiatt (__A, __B, __C); 773132718Skan} 774132718Skan 775132718Skan/* Extract one of the elements of A and sign extend. The selector N must 776132718Skan be immediate. */ 777132718Skan#define _mm_extract_pi8(A, N) __builtin_arm_textrmsb ((__v8qi)(A), (N)) 778132718Skan#define _mm_extract_pi16(A, N) __builtin_arm_textrmsh ((__v4hi)(A), (N)) 779132718Skan#define _mm_extract_pi32(A, N) __builtin_arm_textrmsw ((__v2si)(A), (N)) 780132718Skan 781132718Skan/* Extract one of the elements of A and zero extend. The selector N must 782132718Skan be immediate. */ 783132718Skan#define _mm_extract_pu8(A, N) __builtin_arm_textrmub ((__v8qi)(A), (N)) 784132718Skan#define _mm_extract_pu16(A, N) __builtin_arm_textrmuh ((__v4hi)(A), (N)) 785132718Skan#define _mm_extract_pu32(A, N) __builtin_arm_textrmuw ((__v2si)(A), (N)) 786132718Skan 787132718Skan/* Inserts word D into one of the elements of A. The selector N must be 788132718Skan immediate. */ 789132718Skan#define _mm_insert_pi8(A, D, N) \ 790132718Skan ((__m64) __builtin_arm_tinsrb ((__v8qi)(A), (D), (N))) 791132718Skan#define _mm_insert_pi16(A, D, N) \ 792132718Skan ((__m64) __builtin_arm_tinsrh ((__v4hi)(A), (D), (N))) 793132718Skan#define _mm_insert_pi32(A, D, N) \ 794132718Skan ((__m64) __builtin_arm_tinsrw ((__v2si)(A), (D), (N))) 795132718Skan 796132718Skan/* Compute the element-wise maximum of signed 8-bit values. */ 797132718Skanstatic __inline __m64 798132718Skan_mm_max_pi8 (__m64 __A, __m64 __B) 799132718Skan{ 800132718Skan return (__m64) __builtin_arm_wmaxsb ((__v8qi)__A, (__v8qi)__B); 801132718Skan} 802132718Skan 803132718Skan/* Compute the element-wise maximum of signed 16-bit values. */ 804132718Skanstatic __inline __m64 805132718Skan_mm_max_pi16 (__m64 __A, __m64 __B) 806132718Skan{ 807132718Skan return (__m64) __builtin_arm_wmaxsh ((__v4hi)__A, (__v4hi)__B); 808132718Skan} 809132718Skan 810132718Skan/* Compute the element-wise maximum of signed 32-bit values. */ 811132718Skanstatic __inline __m64 812132718Skan_mm_max_pi32 (__m64 __A, __m64 __B) 813132718Skan{ 814132718Skan return (__m64) __builtin_arm_wmaxsw ((__v2si)__A, (__v2si)__B); 815132718Skan} 816132718Skan 817132718Skan/* Compute the element-wise maximum of unsigned 8-bit values. */ 818132718Skanstatic __inline __m64 819132718Skan_mm_max_pu8 (__m64 __A, __m64 __B) 820132718Skan{ 821132718Skan return (__m64) __builtin_arm_wmaxub ((__v8qi)__A, (__v8qi)__B); 822132718Skan} 823132718Skan 824132718Skan/* Compute the element-wise maximum of unsigned 16-bit values. */ 825132718Skanstatic __inline __m64 826132718Skan_mm_max_pu16 (__m64 __A, __m64 __B) 827132718Skan{ 828132718Skan return (__m64) __builtin_arm_wmaxuh ((__v4hi)__A, (__v4hi)__B); 829132718Skan} 830132718Skan 831132718Skan/* Compute the element-wise maximum of unsigned 32-bit values. */ 832132718Skanstatic __inline __m64 833132718Skan_mm_max_pu32 (__m64 __A, __m64 __B) 834132718Skan{ 835132718Skan return (__m64) __builtin_arm_wmaxuw ((__v2si)__A, (__v2si)__B); 836132718Skan} 837132718Skan 838132718Skan/* Compute the element-wise minimum of signed 16-bit values. */ 839132718Skanstatic __inline __m64 840132718Skan_mm_min_pi8 (__m64 __A, __m64 __B) 841132718Skan{ 842132718Skan return (__m64) __builtin_arm_wminsb ((__v8qi)__A, (__v8qi)__B); 843132718Skan} 844132718Skan 845132718Skan/* Compute the element-wise minimum of signed 16-bit values. */ 846132718Skanstatic __inline __m64 847132718Skan_mm_min_pi16 (__m64 __A, __m64 __B) 848132718Skan{ 849132718Skan return (__m64) __builtin_arm_wminsh ((__v4hi)__A, (__v4hi)__B); 850132718Skan} 851132718Skan 852132718Skan/* Compute the element-wise minimum of signed 32-bit values. */ 853132718Skanstatic __inline __m64 854132718Skan_mm_min_pi32 (__m64 __A, __m64 __B) 855132718Skan{ 856132718Skan return (__m64) __builtin_arm_wminsw ((__v2si)__A, (__v2si)__B); 857132718Skan} 858132718Skan 859132718Skan/* Compute the element-wise minimum of unsigned 16-bit values. */ 860132718Skanstatic __inline __m64 861132718Skan_mm_min_pu8 (__m64 __A, __m64 __B) 862132718Skan{ 863132718Skan return (__m64) __builtin_arm_wminub ((__v8qi)__A, (__v8qi)__B); 864132718Skan} 865132718Skan 866132718Skan/* Compute the element-wise minimum of unsigned 16-bit values. */ 867132718Skanstatic __inline __m64 868132718Skan_mm_min_pu16 (__m64 __A, __m64 __B) 869132718Skan{ 870132718Skan return (__m64) __builtin_arm_wminuh ((__v4hi)__A, (__v4hi)__B); 871132718Skan} 872132718Skan 873132718Skan/* Compute the element-wise minimum of unsigned 32-bit values. */ 874132718Skanstatic __inline __m64 875132718Skan_mm_min_pu32 (__m64 __A, __m64 __B) 876132718Skan{ 877132718Skan return (__m64) __builtin_arm_wminuw ((__v2si)__A, (__v2si)__B); 878132718Skan} 879132718Skan 880132718Skan/* Create an 8-bit mask of the signs of 8-bit values. */ 881132718Skanstatic __inline int 882132718Skan_mm_movemask_pi8 (__m64 __A) 883132718Skan{ 884132718Skan return __builtin_arm_tmovmskb ((__v8qi)__A); 885132718Skan} 886132718Skan 887132718Skan/* Create an 8-bit mask of the signs of 16-bit values. */ 888132718Skanstatic __inline int 889132718Skan_mm_movemask_pi16 (__m64 __A) 890132718Skan{ 891132718Skan return __builtin_arm_tmovmskh ((__v4hi)__A); 892132718Skan} 893132718Skan 894132718Skan/* Create an 8-bit mask of the signs of 32-bit values. */ 895132718Skanstatic __inline int 896132718Skan_mm_movemask_pi32 (__m64 __A) 897132718Skan{ 898132718Skan return __builtin_arm_tmovmskw ((__v2si)__A); 899132718Skan} 900132718Skan 901132718Skan/* Return a combination of the four 16-bit values in A. The selector 902132718Skan must be an immediate. */ 903132718Skan#define _mm_shuffle_pi16(A, N) \ 904132718Skan ((__m64) __builtin_arm_wshufh ((__v4hi)(A), (N))) 905132718Skan 906132718Skan 907132718Skan/* Compute the rounded averages of the unsigned 8-bit values in A and B. */ 908132718Skanstatic __inline __m64 909132718Skan_mm_avg_pu8 (__m64 __A, __m64 __B) 910132718Skan{ 911132718Skan return (__m64) __builtin_arm_wavg2br ((__v8qi)__A, (__v8qi)__B); 912132718Skan} 913132718Skan 914132718Skan/* Compute the rounded averages of the unsigned 16-bit values in A and B. */ 915132718Skanstatic __inline __m64 916132718Skan_mm_avg_pu16 (__m64 __A, __m64 __B) 917132718Skan{ 918132718Skan return (__m64) __builtin_arm_wavg2hr ((__v4hi)__A, (__v4hi)__B); 919132718Skan} 920132718Skan 921132718Skan/* Compute the averages of the unsigned 8-bit values in A and B. */ 922132718Skanstatic __inline __m64 923132718Skan_mm_avg2_pu8 (__m64 __A, __m64 __B) 924132718Skan{ 925132718Skan return (__m64) __builtin_arm_wavg2b ((__v8qi)__A, (__v8qi)__B); 926132718Skan} 927132718Skan 928132718Skan/* Compute the averages of the unsigned 16-bit values in A and B. */ 929132718Skanstatic __inline __m64 930132718Skan_mm_avg2_pu16 (__m64 __A, __m64 __B) 931132718Skan{ 932132718Skan return (__m64) __builtin_arm_wavg2h ((__v4hi)__A, (__v4hi)__B); 933132718Skan} 934132718Skan 935132718Skan/* Compute the sum of the absolute differences of the unsigned 8-bit 936132718Skan values in A and B. Return the value in the lower 16-bit word; the 937132718Skan upper words are cleared. */ 938132718Skanstatic __inline __m64 939132718Skan_mm_sad_pu8 (__m64 __A, __m64 __B) 940132718Skan{ 941132718Skan return (__m64) __builtin_arm_wsadb ((__v8qi)__A, (__v8qi)__B); 942132718Skan} 943132718Skan 944132718Skan/* Compute the sum of the absolute differences of the unsigned 16-bit 945132718Skan values in A and B. Return the value in the lower 32-bit word; the 946132718Skan upper words are cleared. */ 947132718Skanstatic __inline __m64 948132718Skan_mm_sad_pu16 (__m64 __A, __m64 __B) 949132718Skan{ 950132718Skan return (__m64) __builtin_arm_wsadh ((__v4hi)__A, (__v4hi)__B); 951132718Skan} 952132718Skan 953132718Skan/* Compute the sum of the absolute differences of the unsigned 8-bit 954132718Skan values in A and B. Return the value in the lower 16-bit word; the 955132718Skan upper words are cleared. */ 956132718Skanstatic __inline __m64 957132718Skan_mm_sadz_pu8 (__m64 __A, __m64 __B) 958132718Skan{ 959132718Skan return (__m64) __builtin_arm_wsadbz ((__v8qi)__A, (__v8qi)__B); 960132718Skan} 961132718Skan 962132718Skan/* Compute the sum of the absolute differences of the unsigned 16-bit 963132718Skan values in A and B. Return the value in the lower 32-bit word; the 964132718Skan upper words are cleared. */ 965132718Skanstatic __inline __m64 966132718Skan_mm_sadz_pu16 (__m64 __A, __m64 __B) 967132718Skan{ 968132718Skan return (__m64) __builtin_arm_wsadhz ((__v4hi)__A, (__v4hi)__B); 969132718Skan} 970132718Skan 971132718Skanstatic __inline __m64 972132718Skan_mm_align_si64 (__m64 __A, __m64 __B, int __C) 973132718Skan{ 974132718Skan return (__m64) __builtin_arm_walign ((__v8qi)__A, (__v8qi)__B, __C); 975132718Skan} 976132718Skan 977132718Skan/* Creates a 64-bit zero. */ 978132718Skanstatic __inline __m64 979132718Skan_mm_setzero_si64 (void) 980132718Skan{ 981132718Skan return __builtin_arm_wzero (); 982132718Skan} 983132718Skan 984132718Skan/* Set and Get arbitrary iWMMXt Control registers. 985132718Skan Note only registers 0-3 and 8-11 are currently defined, 986132718Skan the rest are reserved. */ 987132718Skan 988132718Skanstatic __inline void 989169689Skan_mm_setwcx (const int __value, const int __regno) 990132718Skan{ 991132718Skan switch (__regno) 992132718Skan { 993169689Skan case 0: __builtin_arm_setwcx (__value, 0); break; 994169689Skan case 1: __builtin_arm_setwcx (__value, 1); break; 995169689Skan case 2: __builtin_arm_setwcx (__value, 2); break; 996169689Skan case 3: __builtin_arm_setwcx (__value, 3); break; 997169689Skan case 8: __builtin_arm_setwcx (__value, 8); break; 998169689Skan case 9: __builtin_arm_setwcx (__value, 9); break; 999169689Skan case 10: __builtin_arm_setwcx (__value, 10); break; 1000169689Skan case 11: __builtin_arm_setwcx (__value, 11); break; 1001132718Skan default: break; 1002132718Skan } 1003132718Skan} 1004132718Skan 1005132718Skanstatic __inline int 1006132718Skan_mm_getwcx (const int __regno) 1007132718Skan{ 1008132718Skan switch (__regno) 1009132718Skan { 1010132718Skan case 0: return __builtin_arm_getwcx (0); 1011132718Skan case 1: return __builtin_arm_getwcx (1); 1012132718Skan case 2: return __builtin_arm_getwcx (2); 1013132718Skan case 3: return __builtin_arm_getwcx (3); 1014132718Skan case 8: return __builtin_arm_getwcx (8); 1015132718Skan case 9: return __builtin_arm_getwcx (9); 1016132718Skan case 10: return __builtin_arm_getwcx (10); 1017132718Skan case 11: return __builtin_arm_getwcx (11); 1018132718Skan default: return 0; 1019132718Skan } 1020132718Skan} 1021132718Skan 1022132718Skan/* Creates a vector of two 32-bit values; I0 is least significant. */ 1023132718Skanstatic __inline __m64 1024132718Skan_mm_set_pi32 (int __i1, int __i0) 1025132718Skan{ 1026132718Skan union { 1027132718Skan __m64 __q; 1028132718Skan struct { 1029132718Skan unsigned int __i0; 1030132718Skan unsigned int __i1; 1031132718Skan } __s; 1032132718Skan } __u; 1033132718Skan 1034132718Skan __u.__s.__i0 = __i0; 1035132718Skan __u.__s.__i1 = __i1; 1036132718Skan 1037132718Skan return __u.__q; 1038132718Skan} 1039132718Skan 1040132718Skan/* Creates a vector of four 16-bit values; W0 is least significant. */ 1041132718Skanstatic __inline __m64 1042132718Skan_mm_set_pi16 (short __w3, short __w2, short __w1, short __w0) 1043132718Skan{ 1044132718Skan unsigned int __i1 = (unsigned short)__w3 << 16 | (unsigned short)__w2; 1045132718Skan unsigned int __i0 = (unsigned short)__w1 << 16 | (unsigned short)__w0; 1046132718Skan return _mm_set_pi32 (__i1, __i0); 1047132718Skan 1048132718Skan} 1049132718Skan 1050132718Skan/* Creates a vector of eight 8-bit values; B0 is least significant. */ 1051132718Skanstatic __inline __m64 1052132718Skan_mm_set_pi8 (char __b7, char __b6, char __b5, char __b4, 1053132718Skan char __b3, char __b2, char __b1, char __b0) 1054132718Skan{ 1055132718Skan unsigned int __i1, __i0; 1056132718Skan 1057132718Skan __i1 = (unsigned char)__b7; 1058132718Skan __i1 = __i1 << 8 | (unsigned char)__b6; 1059132718Skan __i1 = __i1 << 8 | (unsigned char)__b5; 1060132718Skan __i1 = __i1 << 8 | (unsigned char)__b4; 1061132718Skan 1062132718Skan __i0 = (unsigned char)__b3; 1063132718Skan __i0 = __i0 << 8 | (unsigned char)__b2; 1064132718Skan __i0 = __i0 << 8 | (unsigned char)__b1; 1065132718Skan __i0 = __i0 << 8 | (unsigned char)__b0; 1066132718Skan 1067132718Skan return _mm_set_pi32 (__i1, __i0); 1068132718Skan} 1069132718Skan 1070132718Skan/* Similar, but with the arguments in reverse order. */ 1071132718Skanstatic __inline __m64 1072132718Skan_mm_setr_pi32 (int __i0, int __i1) 1073132718Skan{ 1074132718Skan return _mm_set_pi32 (__i1, __i0); 1075132718Skan} 1076132718Skan 1077132718Skanstatic __inline __m64 1078132718Skan_mm_setr_pi16 (short __w0, short __w1, short __w2, short __w3) 1079132718Skan{ 1080132718Skan return _mm_set_pi16 (__w3, __w2, __w1, __w0); 1081132718Skan} 1082132718Skan 1083132718Skanstatic __inline __m64 1084132718Skan_mm_setr_pi8 (char __b0, char __b1, char __b2, char __b3, 1085132718Skan char __b4, char __b5, char __b6, char __b7) 1086132718Skan{ 1087132718Skan return _mm_set_pi8 (__b7, __b6, __b5, __b4, __b3, __b2, __b1, __b0); 1088132718Skan} 1089132718Skan 1090132718Skan/* Creates a vector of two 32-bit values, both elements containing I. */ 1091132718Skanstatic __inline __m64 1092132718Skan_mm_set1_pi32 (int __i) 1093132718Skan{ 1094132718Skan return _mm_set_pi32 (__i, __i); 1095132718Skan} 1096132718Skan 1097132718Skan/* Creates a vector of four 16-bit values, all elements containing W. */ 1098132718Skanstatic __inline __m64 1099132718Skan_mm_set1_pi16 (short __w) 1100132718Skan{ 1101132718Skan unsigned int __i = (unsigned short)__w << 16 | (unsigned short)__w; 1102132718Skan return _mm_set1_pi32 (__i); 1103132718Skan} 1104132718Skan 1105132718Skan/* Creates a vector of four 16-bit values, all elements containing B. */ 1106132718Skanstatic __inline __m64 1107132718Skan_mm_set1_pi8 (char __b) 1108132718Skan{ 1109132718Skan unsigned int __w = (unsigned char)__b << 8 | (unsigned char)__b; 1110132718Skan unsigned int __i = __w << 16 | __w; 1111132718Skan return _mm_set1_pi32 (__i); 1112132718Skan} 1113132718Skan 1114132718Skan/* Convert an integer to a __m64 object. */ 1115132718Skanstatic __inline __m64 1116132718Skan_m_from_int (int __a) 1117132718Skan{ 1118132718Skan return (__m64)__a; 1119132718Skan} 1120132718Skan 1121132718Skan#define _m_packsswb _mm_packs_pi16 1122132718Skan#define _m_packssdw _mm_packs_pi32 1123132718Skan#define _m_packuswb _mm_packs_pu16 1124132718Skan#define _m_packusdw _mm_packs_pu32 1125132718Skan#define _m_packssqd _mm_packs_pi64 1126132718Skan#define _m_packusqd _mm_packs_pu64 1127132718Skan#define _mm_packs_si64 _mm_packs_pi64 1128132718Skan#define _mm_packs_su64 _mm_packs_pu64 1129132718Skan#define _m_punpckhbw _mm_unpackhi_pi8 1130132718Skan#define _m_punpckhwd _mm_unpackhi_pi16 1131132718Skan#define _m_punpckhdq _mm_unpackhi_pi32 1132132718Skan#define _m_punpcklbw _mm_unpacklo_pi8 1133132718Skan#define _m_punpcklwd _mm_unpacklo_pi16 1134132718Skan#define _m_punpckldq _mm_unpacklo_pi32 1135132718Skan#define _m_punpckehsbw _mm_unpackeh_pi8 1136132718Skan#define _m_punpckehswd _mm_unpackeh_pi16 1137132718Skan#define _m_punpckehsdq _mm_unpackeh_pi32 1138132718Skan#define _m_punpckehubw _mm_unpackeh_pu8 1139132718Skan#define _m_punpckehuwd _mm_unpackeh_pu16 1140132718Skan#define _m_punpckehudq _mm_unpackeh_pu32 1141132718Skan#define _m_punpckelsbw _mm_unpackel_pi8 1142132718Skan#define _m_punpckelswd _mm_unpackel_pi16 1143132718Skan#define _m_punpckelsdq _mm_unpackel_pi32 1144132718Skan#define _m_punpckelubw _mm_unpackel_pu8 1145132718Skan#define _m_punpckeluwd _mm_unpackel_pu16 1146132718Skan#define _m_punpckeludq _mm_unpackel_pu32 1147132718Skan#define _m_paddb _mm_add_pi8 1148132718Skan#define _m_paddw _mm_add_pi16 1149132718Skan#define _m_paddd _mm_add_pi32 1150132718Skan#define _m_paddsb _mm_adds_pi8 1151132718Skan#define _m_paddsw _mm_adds_pi16 1152132718Skan#define _m_paddsd _mm_adds_pi32 1153132718Skan#define _m_paddusb _mm_adds_pu8 1154132718Skan#define _m_paddusw _mm_adds_pu16 1155132718Skan#define _m_paddusd _mm_adds_pu32 1156132718Skan#define _m_psubb _mm_sub_pi8 1157132718Skan#define _m_psubw _mm_sub_pi16 1158132718Skan#define _m_psubd _mm_sub_pi32 1159132718Skan#define _m_psubsb _mm_subs_pi8 1160132718Skan#define _m_psubsw _mm_subs_pi16 1161132718Skan#define _m_psubuw _mm_subs_pi32 1162132718Skan#define _m_psubusb _mm_subs_pu8 1163132718Skan#define _m_psubusw _mm_subs_pu16 1164132718Skan#define _m_psubusd _mm_subs_pu32 1165132718Skan#define _m_pmaddwd _mm_madd_pi16 1166132718Skan#define _m_pmadduwd _mm_madd_pu16 1167132718Skan#define _m_pmulhw _mm_mulhi_pi16 1168132718Skan#define _m_pmulhuw _mm_mulhi_pu16 1169132718Skan#define _m_pmullw _mm_mullo_pi16 1170132718Skan#define _m_pmacsw _mm_mac_pi16 1171132718Skan#define _m_pmacuw _mm_mac_pu16 1172132718Skan#define _m_pmacszw _mm_macz_pi16 1173132718Skan#define _m_pmacuzw _mm_macz_pu16 1174132718Skan#define _m_paccb _mm_acc_pu8 1175132718Skan#define _m_paccw _mm_acc_pu16 1176132718Skan#define _m_paccd _mm_acc_pu32 1177132718Skan#define _m_pmia _mm_mia_si64 1178132718Skan#define _m_pmiaph _mm_miaph_si64 1179132718Skan#define _m_pmiabb _mm_miabb_si64 1180132718Skan#define _m_pmiabt _mm_miabt_si64 1181132718Skan#define _m_pmiatb _mm_miatb_si64 1182132718Skan#define _m_pmiatt _mm_miatt_si64 1183132718Skan#define _m_psllw _mm_sll_pi16 1184132718Skan#define _m_psllwi _mm_slli_pi16 1185132718Skan#define _m_pslld _mm_sll_pi32 1186132718Skan#define _m_pslldi _mm_slli_pi32 1187132718Skan#define _m_psllq _mm_sll_si64 1188132718Skan#define _m_psllqi _mm_slli_si64 1189132718Skan#define _m_psraw _mm_sra_pi16 1190132718Skan#define _m_psrawi _mm_srai_pi16 1191132718Skan#define _m_psrad _mm_sra_pi32 1192132718Skan#define _m_psradi _mm_srai_pi32 1193132718Skan#define _m_psraq _mm_sra_si64 1194132718Skan#define _m_psraqi _mm_srai_si64 1195132718Skan#define _m_psrlw _mm_srl_pi16 1196132718Skan#define _m_psrlwi _mm_srli_pi16 1197132718Skan#define _m_psrld _mm_srl_pi32 1198132718Skan#define _m_psrldi _mm_srli_pi32 1199132718Skan#define _m_psrlq _mm_srl_si64 1200132718Skan#define _m_psrlqi _mm_srli_si64 1201132718Skan#define _m_prorw _mm_ror_pi16 1202132718Skan#define _m_prorwi _mm_rori_pi16 1203132718Skan#define _m_prord _mm_ror_pi32 1204132718Skan#define _m_prordi _mm_rori_pi32 1205132718Skan#define _m_prorq _mm_ror_si64 1206132718Skan#define _m_prorqi _mm_rori_si64 1207132718Skan#define _m_pand _mm_and_si64 1208132718Skan#define _m_pandn _mm_andnot_si64 1209132718Skan#define _m_por _mm_or_si64 1210132718Skan#define _m_pxor _mm_xor_si64 1211132718Skan#define _m_pcmpeqb _mm_cmpeq_pi8 1212132718Skan#define _m_pcmpeqw _mm_cmpeq_pi16 1213132718Skan#define _m_pcmpeqd _mm_cmpeq_pi32 1214132718Skan#define _m_pcmpgtb _mm_cmpgt_pi8 1215132718Skan#define _m_pcmpgtub _mm_cmpgt_pu8 1216132718Skan#define _m_pcmpgtw _mm_cmpgt_pi16 1217132718Skan#define _m_pcmpgtuw _mm_cmpgt_pu16 1218132718Skan#define _m_pcmpgtd _mm_cmpgt_pi32 1219132718Skan#define _m_pcmpgtud _mm_cmpgt_pu32 1220132718Skan#define _m_pextrb _mm_extract_pi8 1221132718Skan#define _m_pextrw _mm_extract_pi16 1222132718Skan#define _m_pextrd _mm_extract_pi32 1223132718Skan#define _m_pextrub _mm_extract_pu8 1224132718Skan#define _m_pextruw _mm_extract_pu16 1225132718Skan#define _m_pextrud _mm_extract_pu32 1226132718Skan#define _m_pinsrb _mm_insert_pi8 1227132718Skan#define _m_pinsrw _mm_insert_pi16 1228132718Skan#define _m_pinsrd _mm_insert_pi32 1229132718Skan#define _m_pmaxsb _mm_max_pi8 1230132718Skan#define _m_pmaxsw _mm_max_pi16 1231132718Skan#define _m_pmaxsd _mm_max_pi32 1232132718Skan#define _m_pmaxub _mm_max_pu8 1233132718Skan#define _m_pmaxuw _mm_max_pu16 1234132718Skan#define _m_pmaxud _mm_max_pu32 1235132718Skan#define _m_pminsb _mm_min_pi8 1236132718Skan#define _m_pminsw _mm_min_pi16 1237132718Skan#define _m_pminsd _mm_min_pi32 1238132718Skan#define _m_pminub _mm_min_pu8 1239132718Skan#define _m_pminuw _mm_min_pu16 1240132718Skan#define _m_pminud _mm_min_pu32 1241132718Skan#define _m_pmovmskb _mm_movemask_pi8 1242132718Skan#define _m_pmovmskw _mm_movemask_pi16 1243132718Skan#define _m_pmovmskd _mm_movemask_pi32 1244132718Skan#define _m_pshufw _mm_shuffle_pi16 1245132718Skan#define _m_pavgb _mm_avg_pu8 1246132718Skan#define _m_pavgw _mm_avg_pu16 1247132718Skan#define _m_pavg2b _mm_avg2_pu8 1248132718Skan#define _m_pavg2w _mm_avg2_pu16 1249132718Skan#define _m_psadbw _mm_sad_pu8 1250132718Skan#define _m_psadwd _mm_sad_pu16 1251132718Skan#define _m_psadzbw _mm_sadz_pu8 1252132718Skan#define _m_psadzwd _mm_sadz_pu16 1253132718Skan#define _m_paligniq _mm_align_si64 1254132718Skan#define _m_cvt_si2pi _mm_cvtsi64_m64 1255132718Skan#define _m_cvt_pi2si _mm_cvtm64_si64 1256132718Skan 1257132718Skan#endif /* _MMINTRIN_H_INCLUDED */ 1258