v64_intrinsics_c.h 22.6 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93
/*
 * Copyright (c) 2016, Alliance for Open Media. All rights reserved
 *
 * This source code is subject to the terms of the BSD 2 Clause License and
 * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
 * was not distributed with this source code in the LICENSE file, you can
 * obtain it at www.aomedia.org/license/software. If the Alliance for Open
 * Media Patent License 1.0 was not distributed with this source code in the
 * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
 */

#ifndef _V64_INTRINSICS_C_H
#define _V64_INTRINSICS_C_H

/* Note: This implements the intrinsics in plain, unoptimised C.
   Intended for reference, porting or debugging. */

#include <stdio.h>
#include <stdlib.h>
#include "./aom_config.h"

typedef union {
  uint8_t u8[8];
  uint16_t u16[4];
  uint32_t u32[2];
  uint64_t u64;
  int8_t s8[8];
  int16_t s16[4];
  int32_t s32[2];
  int64_t s64;
} c_v64;

SIMD_INLINE uint32_t c_v64_low_u32(c_v64 a) { return a.u32[CONFIG_BIG_ENDIAN]; }

SIMD_INLINE uint32_t c_v64_high_u32(c_v64 a) {
  return a.u32[!CONFIG_BIG_ENDIAN];
}

SIMD_INLINE int32_t c_v64_low_s32(c_v64 a) { return a.s32[CONFIG_BIG_ENDIAN]; }

SIMD_INLINE int32_t c_v64_high_s32(c_v64 a) {
  return a.s32[!CONFIG_BIG_ENDIAN];
}

SIMD_INLINE c_v64 c_v64_from_32(uint32_t x, uint32_t y) {
  c_v64 t;
  t.u32[!CONFIG_BIG_ENDIAN] = x;
  t.u32[CONFIG_BIG_ENDIAN] = y;
  return t;
}

SIMD_INLINE c_v64 c_v64_from_64(uint64_t x) {
  c_v64 t;
  t.u64 = x;
  return t;
}

SIMD_INLINE uint64_t c_v64_u64(c_v64 x) { return x.u64; }

SIMD_INLINE c_v64 c_v64_from_16(uint16_t a, uint16_t b, uint16_t c,
                                uint16_t d) {
  c_v64 t;
  if (CONFIG_BIG_ENDIAN) {
    t.u16[0] = a;
    t.u16[1] = b;
    t.u16[2] = c;
    t.u16[3] = d;
  } else {
    t.u16[3] = a;
    t.u16[2] = b;
    t.u16[1] = c;
    t.u16[0] = d;
  }
  return t;
}

SIMD_INLINE uint32_t c_u32_load_unaligned(const void *p) {
  uint32_t t;
  uint8_t *pp = (uint8_t *)p;
  uint8_t *q = (uint8_t *)&t;
  int c;
  for (c = 0; c < 4; c++) q[c] = pp[c];
  return t;
}

SIMD_INLINE void c_u32_store_unaligned(void *p, uint32_t a) {
  uint8_t *pp = (uint8_t *)p;
  uint8_t *q = (uint8_t *)&a;
  int c;
  for (c = 0; c < 4; c++) pp[c] = q[c];
}

SIMD_INLINE uint32_t c_u32_load_aligned(const void *p) {
94
  if (SIMD_CHECK && (uintptr_t)p & 3) {
95 96 97 98 99 100 101
    fprintf(stderr, "Error: Unaligned u32 load at %p\n", p);
    abort();
  }
  return c_u32_load_unaligned(p);
}

SIMD_INLINE void c_u32_store_aligned(void *p, uint32_t a) {
102
  if (SIMD_CHECK && (uintptr_t)p & 3) {
103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118
    fprintf(stderr, "Error: Unaligned u32 store at %p\n", p);
    abort();
  }
  c_u32_store_unaligned(p, a);
}

SIMD_INLINE c_v64 c_v64_load_unaligned(const void *p) {
  c_v64 t;
  uint8_t *pp = (uint8_t *)p;
  uint8_t *q = (uint8_t *)&t;
  int c;
  for (c = 0; c < 8; c++) q[c] = pp[c];
  return t;
}

SIMD_INLINE c_v64 c_v64_load_aligned(const void *p) {
119
  if (SIMD_CHECK && (uintptr_t)p & 7) {
120 121 122 123 124 125 126 127 128 129 130 131 132 133
    fprintf(stderr, "Error: Unaligned c_v64 load at %p\n", p);
    abort();
  }
  return c_v64_load_unaligned(p);
}

SIMD_INLINE void c_v64_store_unaligned(void *p, c_v64 a) {
  uint8_t *q = (uint8_t *)p;
  uint8_t *r = (uint8_t *)&a;
  int c;
  for (c = 0; c < 8; c++) q[c] = r[c];
}

SIMD_INLINE void c_v64_store_aligned(void *p, c_v64 a) {
134
  if (SIMD_CHECK && (uintptr_t)p & 7) {
135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193
    fprintf(stderr, "Error: Unaligned c_v64 store at %p\n", p);
    abort();
  }
  c_v64_store_unaligned(p, a);
}

SIMD_INLINE c_v64 c_v64_zero() {
  c_v64 t;
  t.u64 = 0;
  return t;
}

SIMD_INLINE c_v64 c_v64_dup_8(uint8_t x) {
  c_v64 t;
  t.u8[0] = t.u8[1] = t.u8[2] = t.u8[3] = t.u8[4] = t.u8[5] = t.u8[6] =
      t.u8[7] = x;
  return t;
}

SIMD_INLINE c_v64 c_v64_dup_16(uint16_t x) {
  c_v64 t;
  t.u16[0] = t.u16[1] = t.u16[2] = t.u16[3] = x;
  return t;
}

SIMD_INLINE c_v64 c_v64_dup_32(uint32_t x) {
  c_v64 t;
  t.u32[0] = t.u32[1] = x;
  return t;
}

SIMD_INLINE c_v64 c_v64_add_8(c_v64 a, c_v64 b) {
  c_v64 t;
  int c;
  for (c = 0; c < 8; c++) t.u8[c] = a.u8[c] + b.u8[c];
  return t;
}

SIMD_INLINE c_v64 c_v64_add_16(c_v64 a, c_v64 b) {
  c_v64 t;
  int c;
  for (c = 0; c < 4; c++) t.u16[c] = a.u16[c] + b.u16[c];
  return t;
}

SIMD_INLINE c_v64 c_v64_sadd_s16(c_v64 a, c_v64 b) {
  c_v64 t;
  int c;
  for (c = 0; c < 4; c++)
    t.s16[c] = (int32_t)a.s16[c] + (int32_t)b.s16[c] > 32767
                   ? 32767
                   : (int32_t)a.s16[c] + (int32_t)b.s16[c] < -32768
                         ? -32768
                         : (int32_t)a.s16[c] + (int32_t)b.s16[c];
  return t;
}

SIMD_INLINE c_v64 c_v64_add_32(c_v64 a, c_v64 b) {
  c_v64 t;
194 195
  t.u32[0] = (uint32_t)((uint64_t)a.u32[0] + b.u32[0]);
  t.u32[1] = (uint32_t)((uint64_t)a.u32[1] + b.u32[1]);
196 197 198 199 200 201 202 203 204 205 206 207 208 209
  return t;
}

SIMD_INLINE c_v64 c_v64_sub_8(c_v64 a, c_v64 b) {
  c_v64 t;
  int c;
  for (c = 0; c < 8; c++) t.u8[c] = a.u8[c] - b.u8[c];
  return t;
}

SIMD_INLINE c_v64 c_v64_ssub_u8(c_v64 a, c_v64 b) {
  c_v64 t;
  int c;
  for (c = 0; c < 8; c++)
210
    t.u8[c] = (int32_t)a.u8[c] - (int32_t)b.u8[c] < 0 ? 0 : a.u8[c] - b.u8[c];
211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242
  return t;
}

SIMD_INLINE c_v64 c_v64_ssub_s8(c_v64 a, c_v64 b) {
  c_v64 t;
  int c;
  for (c = 0; c < 8; c++) {
    int16_t d = (int16_t)a.s8[c] - (int16_t)b.s8[c];
    t.s8[c] = d > 127 ? 127 : (d < -128 ? -128 : d);
  }
  return t;
}

SIMD_INLINE c_v64 c_v64_sub_16(c_v64 a, c_v64 b) {
  c_v64 t;
  int c;
  for (c = 0; c < 4; c++) t.u16[c] = a.u16[c] - b.u16[c];
  return t;
}

SIMD_INLINE c_v64 c_v64_ssub_s16(c_v64 a, c_v64 b) {
  c_v64 t;
  int c;
  for (c = 0; c < 4; c++)
    t.s16[c] = (int32_t)a.s16[c] - (int32_t)b.s16[c] < -32768
                   ? -32768
                   : (int32_t)a.s16[c] - (int32_t)b.s16[c] > 32767
                         ? 32767
                         : (int32_t)a.s16[c] - (int32_t)b.s16[c];
  return t;
}

243 244 245 246 247 248 249 250 251
SIMD_INLINE c_v64 c_v64_ssub_u16(c_v64 a, c_v64 b) {
  c_v64 t;
  int c;
  for (c = 0; c < 4; c++)
    t.u16[c] =
        (int32_t)a.u16[c] - (int32_t)b.u16[c] < 0 ? 0 : a.u16[c] - b.u16[c];
  return t;
}

252 253
SIMD_INLINE c_v64 c_v64_sub_32(c_v64 a, c_v64 b) {
  c_v64 t;
254 255
  t.u32[0] = (uint32_t)((int64_t)a.u32[0] - b.u32[0]);
  t.u32[1] = (uint32_t)((int64_t)a.u32[1] - b.u32[1]);
256 257 258 259 260 261 262 263 264 265 266
  return t;
}

SIMD_INLINE c_v64 c_v64_abs_s16(c_v64 a) {
  c_v64 t;
  int c;
  for (c = 0; c < 4; c++)
    t.u16[c] = (int16_t)a.u16[c] > 0 ? a.u16[c] : -a.u16[c];
  return t;
}

267 268 269 270 271 272 273
SIMD_INLINE c_v64 c_v64_abs_s8(c_v64 a) {
  c_v64 t;
  int c;
  for (c = 0; c < 8; c++) t.u8[c] = (int8_t)a.u8[c] > 0 ? a.u8[c] : -a.u8[c];
  return t;
}

274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427
SIMD_INLINE c_v64 _c_v64_zip_8(c_v64 a, c_v64 b, int mode) {
  c_v64 t;
  if (mode) {
    t.u8[7] = a.u8[7];
    t.u8[6] = b.u8[7];
    t.u8[5] = a.u8[6];
    t.u8[4] = b.u8[6];
    t.u8[3] = a.u8[5];
    t.u8[2] = b.u8[5];
    t.u8[1] = a.u8[4];
    t.u8[0] = b.u8[4];
  } else {
    t.u8[7] = a.u8[3];
    t.u8[6] = b.u8[3];
    t.u8[5] = a.u8[2];
    t.u8[4] = b.u8[2];
    t.u8[3] = a.u8[1];
    t.u8[2] = b.u8[1];
    t.u8[1] = a.u8[0];
    t.u8[0] = b.u8[0];
  }
  return t;
}

SIMD_INLINE c_v64 c_v64_ziplo_8(c_v64 a, c_v64 b) {
  return CONFIG_BIG_ENDIAN ? _c_v64_zip_8(b, a, 1) : _c_v64_zip_8(a, b, 0);
}

SIMD_INLINE c_v64 c_v64_ziphi_8(c_v64 a, c_v64 b) {
  return CONFIG_BIG_ENDIAN ? _c_v64_zip_8(b, a, 0) : _c_v64_zip_8(a, b, 1);
}

SIMD_INLINE c_v64 _c_v64_zip_16(c_v64 a, c_v64 b, int mode) {
  c_v64 t;
  if (mode) {
    t.u16[3] = a.u16[3];
    t.u16[2] = b.u16[3];
    t.u16[1] = a.u16[2];
    t.u16[0] = b.u16[2];
  } else {
    t.u16[3] = a.u16[1];
    t.u16[2] = b.u16[1];
    t.u16[1] = a.u16[0];
    t.u16[0] = b.u16[0];
  }
  return t;
}

SIMD_INLINE c_v64 c_v64_ziplo_16(c_v64 a, c_v64 b) {
  return CONFIG_BIG_ENDIAN ? _c_v64_zip_16(b, a, 1) : _c_v64_zip_16(a, b, 0);
}

SIMD_INLINE c_v64 c_v64_ziphi_16(c_v64 a, c_v64 b) {
  return CONFIG_BIG_ENDIAN ? _c_v64_zip_16(b, a, 0) : _c_v64_zip_16(a, b, 1);
}

SIMD_INLINE c_v64 _c_v64_zip_32(c_v64 a, c_v64 b, int mode) {
  c_v64 t;
  if (mode) {
    t.u32[1] = a.u32[1];
    t.u32[0] = b.u32[1];
  } else {
    t.u32[1] = a.u32[0];
    t.u32[0] = b.u32[0];
  }
  return t;
}

SIMD_INLINE c_v64 c_v64_ziplo_32(c_v64 a, c_v64 b) {
  return CONFIG_BIG_ENDIAN ? _c_v64_zip_32(b, a, 1) : _c_v64_zip_32(a, b, 0);
}

SIMD_INLINE c_v64 c_v64_ziphi_32(c_v64 a, c_v64 b) {
  return CONFIG_BIG_ENDIAN ? _c_v64_zip_32(b, a, 0) : _c_v64_zip_32(a, b, 1);
}

SIMD_INLINE c_v64 _c_v64_unzip_8(c_v64 a, c_v64 b, int mode) {
  c_v64 t;
  if (mode) {
    t.u8[7] = b.u8[7];
    t.u8[6] = b.u8[5];
    t.u8[5] = b.u8[3];
    t.u8[4] = b.u8[1];
    t.u8[3] = a.u8[7];
    t.u8[2] = a.u8[5];
    t.u8[1] = a.u8[3];
    t.u8[0] = a.u8[1];
  } else {
    t.u8[7] = a.u8[6];
    t.u8[6] = a.u8[4];
    t.u8[5] = a.u8[2];
    t.u8[4] = a.u8[0];
    t.u8[3] = b.u8[6];
    t.u8[2] = b.u8[4];
    t.u8[1] = b.u8[2];
    t.u8[0] = b.u8[0];
  }
  return t;
}

SIMD_INLINE c_v64 c_v64_unziplo_8(c_v64 a, c_v64 b) {
  return CONFIG_BIG_ENDIAN ? _c_v64_unzip_8(a, b, 1) : _c_v64_unzip_8(a, b, 0);
}

SIMD_INLINE c_v64 c_v64_unziphi_8(c_v64 a, c_v64 b) {
  return CONFIG_BIG_ENDIAN ? _c_v64_unzip_8(b, a, 0) : _c_v64_unzip_8(b, a, 1);
}

SIMD_INLINE c_v64 _c_v64_unzip_16(c_v64 a, c_v64 b, int mode) {
  c_v64 t;
  if (mode) {
    t.u16[3] = b.u16[3];
    t.u16[2] = b.u16[1];
    t.u16[1] = a.u16[3];
    t.u16[0] = a.u16[1];
  } else {
    t.u16[3] = a.u16[2];
    t.u16[2] = a.u16[0];
    t.u16[1] = b.u16[2];
    t.u16[0] = b.u16[0];
  }
  return t;
}

SIMD_INLINE c_v64 c_v64_unziplo_16(c_v64 a, c_v64 b) {
  return CONFIG_BIG_ENDIAN ? _c_v64_unzip_16(a, b, 1)
                           : _c_v64_unzip_16(a, b, 0);
}

SIMD_INLINE c_v64 c_v64_unziphi_16(c_v64 a, c_v64 b) {
  return CONFIG_BIG_ENDIAN ? _c_v64_unzip_16(b, a, 0)
                           : _c_v64_unzip_16(b, a, 1);
}

SIMD_INLINE c_v64 c_v64_unpacklo_u8_s16(c_v64 a) {
  c_v64 t;
  int endian = !!CONFIG_BIG_ENDIAN * 4;
  t.s16[3] = (int16_t)a.u8[3 + endian];
  t.s16[2] = (int16_t)a.u8[2 + endian];
  t.s16[1] = (int16_t)a.u8[1 + endian];
  t.s16[0] = (int16_t)a.u8[0 + endian];
  return t;
}

SIMD_INLINE c_v64 c_v64_unpackhi_u8_s16(c_v64 a) {
  c_v64 t;
  int endian = !!CONFIG_BIG_ENDIAN * 4;
  t.s16[3] = (int16_t)a.u8[7 - endian];
  t.s16[2] = (int16_t)a.u8[6 - endian];
  t.s16[1] = (int16_t)a.u8[5 - endian];
  t.s16[0] = (int16_t)a.u8[4 - endian];
  return t;
}

428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447
SIMD_INLINE c_v64 c_v64_unpacklo_s8_s16(c_v64 a) {
  c_v64 t;
  int endian = !!CONFIG_BIG_ENDIAN * 4;
  t.s16[3] = (int16_t)a.s8[3 + endian];
  t.s16[2] = (int16_t)a.s8[2 + endian];
  t.s16[1] = (int16_t)a.s8[1 + endian];
  t.s16[0] = (int16_t)a.s8[0 + endian];
  return t;
}

SIMD_INLINE c_v64 c_v64_unpackhi_s8_s16(c_v64 a) {
  c_v64 t;
  int endian = !!CONFIG_BIG_ENDIAN * 4;
  t.s16[3] = (int16_t)a.s8[7 - endian];
  t.s16[2] = (int16_t)a.s8[6 - endian];
  t.s16[1] = (int16_t)a.s8[5 - endian];
  t.s16[0] = (int16_t)a.s8[4 - endian];
  return t;
}

448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529
SIMD_INLINE c_v64 c_v64_pack_s32_s16(c_v64 a, c_v64 b) {
  c_v64 t;
  if (CONFIG_BIG_ENDIAN) {
    c_v64 u = a;
    a = b;
    b = u;
  }
  t.s16[3] = a.s32[1] > 32767 ? 32767 : a.s32[1] < -32768 ? -32768 : a.s32[1];
  t.s16[2] = a.s32[0] > 32767 ? 32767 : a.s32[0] < -32768 ? -32768 : a.s32[0];
  t.s16[1] = b.s32[1] > 32767 ? 32767 : b.s32[1] < -32768 ? -32768 : b.s32[1];
  t.s16[0] = b.s32[0] > 32767 ? 32767 : b.s32[0] < -32768 ? -32768 : b.s32[0];
  return t;
}

SIMD_INLINE c_v64 c_v64_pack_s16_u8(c_v64 a, c_v64 b) {
  c_v64 t;
  if (CONFIG_BIG_ENDIAN) {
    c_v64 u = a;
    a = b;
    b = u;
  }
  t.u8[7] = a.s16[3] > 255 ? 255 : a.s16[3] < 0 ? 0 : a.s16[3];
  t.u8[6] = a.s16[2] > 255 ? 255 : a.s16[2] < 0 ? 0 : a.s16[2];
  t.u8[5] = a.s16[1] > 255 ? 255 : a.s16[1] < 0 ? 0 : a.s16[1];
  t.u8[4] = a.s16[0] > 255 ? 255 : a.s16[0] < 0 ? 0 : a.s16[0];
  t.u8[3] = b.s16[3] > 255 ? 255 : b.s16[3] < 0 ? 0 : b.s16[3];
  t.u8[2] = b.s16[2] > 255 ? 255 : b.s16[2] < 0 ? 0 : b.s16[2];
  t.u8[1] = b.s16[1] > 255 ? 255 : b.s16[1] < 0 ? 0 : b.s16[1];
  t.u8[0] = b.s16[0] > 255 ? 255 : b.s16[0] < 0 ? 0 : b.s16[0];
  return t;
}

SIMD_INLINE c_v64 c_v64_pack_s16_s8(c_v64 a, c_v64 b) {
  c_v64 t;
  if (CONFIG_BIG_ENDIAN) {
    c_v64 u = a;
    a = b;
    b = u;
  }
  t.u8[7] = a.s16[3] > 127 ? 127 : a.s16[3] < -128 ? 128 : a.s16[3];
  t.u8[6] = a.s16[2] > 127 ? 127 : a.s16[2] < -128 ? 128 : a.s16[2];
  t.u8[5] = a.s16[1] > 127 ? 127 : a.s16[1] < -128 ? 128 : a.s16[1];
  t.u8[4] = a.s16[0] > 127 ? 127 : a.s16[0] < -128 ? 128 : a.s16[0];
  t.u8[3] = b.s16[3] > 127 ? 127 : b.s16[3] < -128 ? 128 : b.s16[3];
  t.u8[2] = b.s16[2] > 127 ? 127 : b.s16[2] < -128 ? 128 : b.s16[2];
  t.u8[1] = b.s16[1] > 127 ? 127 : b.s16[1] < -128 ? 128 : b.s16[1];
  t.u8[0] = b.s16[0] > 127 ? 127 : b.s16[0] < -128 ? 128 : b.s16[0];
  return t;
}

SIMD_INLINE c_v64 c_v64_unpacklo_u16_s32(c_v64 a) {
  c_v64 t;
  t.s32[1] = a.u16[1 + !!CONFIG_BIG_ENDIAN * 2];
  t.s32[0] = a.u16[0 + !!CONFIG_BIG_ENDIAN * 2];
  return t;
}

SIMD_INLINE c_v64 c_v64_unpacklo_s16_s32(c_v64 a) {
  c_v64 t;
  t.s32[1] = a.s16[1 + !!CONFIG_BIG_ENDIAN * 2];
  t.s32[0] = a.s16[0 + !!CONFIG_BIG_ENDIAN * 2];
  return t;
}

SIMD_INLINE c_v64 c_v64_unpackhi_u16_s32(c_v64 a) {
  c_v64 t;
  t.s32[1] = a.u16[3 - !!CONFIG_BIG_ENDIAN * 2];
  t.s32[0] = a.u16[2 - !!CONFIG_BIG_ENDIAN * 2];
  return t;
}

SIMD_INLINE c_v64 c_v64_unpackhi_s16_s32(c_v64 a) {
  c_v64 t;
  t.s32[1] = a.s16[3 - !!CONFIG_BIG_ENDIAN * 2];
  t.s32[0] = a.s16[2 - !!CONFIG_BIG_ENDIAN * 2];
  return t;
}

SIMD_INLINE c_v64 c_v64_shuffle_8(c_v64 a, c_v64 pattern) {
  c_v64 t;
  int c;
  for (c = 0; c < 8; c++) {
530
    if (SIMD_CHECK && (pattern.u8[c] & ~7)) {
531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632
      fprintf(stderr, "Error: Undefined v64_shuffle_8 index %d/%d\n",
              pattern.u8[c], c);
      abort();
    }
    t.u8[c] =
        a.u8[CONFIG_BIG_ENDIAN ? 7 - (pattern.u8[c] & 7) : pattern.u8[c] & 7];
  }
  return t;
}

SIMD_INLINE int64_t c_v64_dotp_su8(c_v64 a, c_v64 b) {
  return a.s8[7] * b.u8[7] + a.s8[6] * b.u8[6] + a.s8[5] * b.u8[5] +
         a.s8[4] * b.u8[4] + a.s8[3] * b.u8[3] + a.s8[2] * b.u8[2] +
         a.s8[1] * b.u8[1] + a.s8[0] * b.u8[0];
}

SIMD_INLINE int64_t c_v64_dotp_s16(c_v64 a, c_v64 b) {
  return (int64_t)(a.s16[3] * b.s16[3] + a.s16[2] * b.s16[2]) +
         (int64_t)(a.s16[1] * b.s16[1] + a.s16[0] * b.s16[0]);
}

SIMD_INLINE uint64_t c_v64_hadd_u8(c_v64 a) {
  return a.u8[7] + a.u8[6] + a.u8[5] + a.u8[4] + a.u8[3] + a.u8[2] + a.u8[1] +
         a.u8[0];
}

SIMD_INLINE int64_t c_v64_hadd_s16(c_v64 a) {
  return a.s16[3] + a.s16[2] + a.s16[1] + a.s16[0];
}

typedef uint32_t c_sad64_internal;

/* Implementation dependent return value.  Result must be finalised with
   v64_sad_u8_sum().
   The result for more than 32 v64_sad_u8() calls is undefined. */
SIMD_INLINE c_sad64_internal c_v64_sad_u8_init() { return 0; }

SIMD_INLINE c_sad64_internal c_v64_sad_u8(c_sad64_internal s, c_v64 a,
                                          c_v64 b) {
  int c;
  for (c = 0; c < 8; c++)
    s += a.u8[c] > b.u8[c] ? a.u8[c] - b.u8[c] : b.u8[c] - a.u8[c];
  return s;
}

SIMD_INLINE uint32_t c_v64_sad_u8_sum(c_sad64_internal s) { return s; }

typedef uint32_t c_ssd64_internal;

/* Implementation dependent return value.  Result must be finalised with
 * v64_ssd_u8_sum(). */
SIMD_INLINE c_ssd64_internal c_v64_ssd_u8_init() { return 0; }

SIMD_INLINE c_ssd64_internal c_v64_ssd_u8(c_ssd64_internal s, c_v64 a,
                                          c_v64 b) {
  int c;
  for (c = 0; c < 8; c++) s += (a.u8[c] - b.u8[c]) * (a.u8[c] - b.u8[c]);
  return s;
}

SIMD_INLINE uint32_t c_v64_ssd_u8_sum(c_ssd64_internal s) { return s; }

SIMD_INLINE c_v64 c_v64_or(c_v64 a, c_v64 b) {
  c_v64 t;
  t.u64 = a.u64 | b.u64;
  return t;
}

SIMD_INLINE c_v64 c_v64_xor(c_v64 a, c_v64 b) {
  c_v64 t;
  t.u64 = a.u64 ^ b.u64;
  return t;
}

SIMD_INLINE c_v64 c_v64_and(c_v64 a, c_v64 b) {
  c_v64 t;
  t.u64 = a.u64 & b.u64;
  return t;
}

SIMD_INLINE c_v64 c_v64_andn(c_v64 a, c_v64 b) {
  c_v64 t;
  t.u64 = a.u64 & ~b.u64;
  return t;
}

SIMD_INLINE c_v64 c_v64_mullo_s16(c_v64 a, c_v64 b) {
  c_v64 t;
  int c;
  for (c = 0; c < 4; c++) t.s16[c] = (int16_t)(a.s16[c] * b.s16[c]);
  return t;
}

SIMD_INLINE c_v64 c_v64_mulhi_s16(c_v64 a, c_v64 b) {
  c_v64 t;
  int c;
  for (c = 0; c < 4; c++) t.s16[c] = (a.s16[c] * b.s16[c]) >> 16;
  return t;
}

SIMD_INLINE c_v64 c_v64_mullo_s32(c_v64 a, c_v64 b) {
  c_v64 t;
Steinar Midtskogen's avatar
Steinar Midtskogen committed
633 634
  t.s32[0] = (int32_t)((int64_t)a.s32[0] * b.s32[0]);
  t.s32[1] = (int32_t)((int64_t)a.s32[1] * b.s32[1]);
635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766
  return t;
}

SIMD_INLINE c_v64 c_v64_madd_s16(c_v64 a, c_v64 b) {
  c_v64 t;
  t.s32[0] = a.s16[0] * b.s16[0] + a.s16[1] * b.s16[1];
  t.s32[1] = a.s16[2] * b.s16[2] + a.s16[3] * b.s16[3];
  return t;
}

SIMD_INLINE c_v64 c_v64_madd_us8(c_v64 a, c_v64 b) {
  c_v64 t;
  int32_t u;
  u = a.u8[0] * b.s8[0] + a.u8[1] * b.s8[1];
  t.s16[0] = u > 32767 ? 32767 : u < -32768 ? -32768 : u;
  u = a.u8[2] * b.s8[2] + a.u8[3] * b.s8[3];
  t.s16[1] = u > 32767 ? 32767 : u < -32768 ? -32768 : u;
  u = a.u8[4] * b.s8[4] + a.u8[5] * b.s8[5];
  t.s16[2] = u > 32767 ? 32767 : u < -32768 ? -32768 : u;
  u = a.u8[6] * b.s8[6] + a.u8[7] * b.s8[7];
  t.s16[3] = u > 32767 ? 32767 : u < -32768 ? -32768 : u;
  return t;
}

SIMD_INLINE c_v64 c_v64_avg_u8(c_v64 a, c_v64 b) {
  c_v64 t;
  int c;
  for (c = 0; c < 8; c++) t.u8[c] = (a.u8[c] + b.u8[c] + 1) >> 1;
  return t;
}

SIMD_INLINE c_v64 c_v64_rdavg_u8(c_v64 a, c_v64 b) {
  c_v64 t;
  int c;
  for (c = 0; c < 8; c++) t.u8[c] = (a.u8[c] + b.u8[c]) >> 1;
  return t;
}

SIMD_INLINE c_v64 c_v64_avg_u16(c_v64 a, c_v64 b) {
  c_v64 t;
  int c;
  for (c = 0; c < 4; c++) t.u16[c] = (a.u16[c] + b.u16[c] + 1) >> 1;
  return t;
}

SIMD_INLINE c_v64 c_v64_min_u8(c_v64 a, c_v64 b) {
  c_v64 t;
  int c;
  for (c = 0; c < 8; c++) t.u8[c] = a.u8[c] > b.u8[c] ? b.u8[c] : a.u8[c];
  return t;
}

SIMD_INLINE c_v64 c_v64_max_u8(c_v64 a, c_v64 b) {
  c_v64 t;
  int c;
  for (c = 0; c < 8; c++) t.u8[c] = a.u8[c] > b.u8[c] ? a.u8[c] : b.u8[c];
  return t;
}

SIMD_INLINE c_v64 c_v64_min_s8(c_v64 a, c_v64 b) {
  c_v64 t;
  int c;
  for (c = 0; c < 8; c++) t.s8[c] = a.s8[c] > b.s8[c] ? b.s8[c] : a.s8[c];
  return t;
}

SIMD_INLINE c_v64 c_v64_max_s8(c_v64 a, c_v64 b) {
  c_v64 t;
  int c;
  for (c = 0; c < 8; c++) t.s8[c] = a.s8[c] > b.s8[c] ? a.s8[c] : b.s8[c];
  return t;
}

SIMD_INLINE c_v64 c_v64_min_s16(c_v64 a, c_v64 b) {
  c_v64 t;
  int c;
  for (c = 0; c < 4; c++) t.s16[c] = a.s16[c] > b.s16[c] ? b.s16[c] : a.s16[c];
  return t;
}

SIMD_INLINE c_v64 c_v64_max_s16(c_v64 a, c_v64 b) {
  c_v64 t;
  int c;
  for (c = 0; c < 4; c++) t.s16[c] = a.s16[c] > b.s16[c] ? a.s16[c] : b.s16[c];
  return t;
}

SIMD_INLINE c_v64 c_v64_cmpgt_s8(c_v64 a, c_v64 b) {
  c_v64 t;
  int c;
  for (c = 0; c < 8; c++) t.s8[c] = -(a.s8[c] > b.s8[c]);
  return t;
}

SIMD_INLINE c_v64 c_v64_cmplt_s8(c_v64 a, c_v64 b) {
  c_v64 t;
  int c;
  for (c = 0; c < 8; c++) t.s8[c] = -(a.s8[c] < b.s8[c]);
  return t;
}

SIMD_INLINE c_v64 c_v64_cmpeq_8(c_v64 a, c_v64 b) {
  c_v64 t;
  int c;
  for (c = 0; c < 8; c++) t.s8[c] = -(a.u8[c] == b.u8[c]);
  return t;
}

SIMD_INLINE c_v64 c_v64_cmpgt_s16(c_v64 a, c_v64 b) {
  c_v64 t;
  int c;
  for (c = 0; c < 4; c++) t.s16[c] = -(a.s16[c] > b.s16[c]);
  return t;
}

SIMD_INLINE c_v64 c_v64_cmplt_s16(c_v64 a, c_v64 b) {
  c_v64 t;
  int c;
  for (c = 0; c < 4; c++) t.s16[c] = -(a.s16[c] < b.s16[c]);
  return t;
}

SIMD_INLINE c_v64 c_v64_cmpeq_16(c_v64 a, c_v64 b) {
  c_v64 t;
  int c;
  for (c = 0; c < 4; c++) t.s16[c] = -(a.u16[c] == b.u16[c]);
  return t;
}

SIMD_INLINE c_v64 c_v64_shl_8(c_v64 a, unsigned int n) {
  c_v64 t;
  int c;
767
  if (SIMD_CHECK && n > 7) {
768 769 770 771 772 773 774 775 776 777
    fprintf(stderr, "Error: Undefined u8 shift left %d\n", n);
    abort();
  }
  for (c = 0; c < 8; c++) t.s8[c] = a.u8[c] << n;
  return t;
}

SIMD_INLINE c_v64 c_v64_shr_u8(c_v64 a, unsigned int n) {
  c_v64 t;
  int c;
778
  if (SIMD_CHECK && n > 7) {
779 780 781 782 783 784 785 786 787 788
    fprintf(stderr, "Error: Undefined u8 shift right %d\n", n);
    abort();
  }
  for (c = 0; c < 8; c++) t.u8[c] = a.u8[c] >> n;
  return t;
}

SIMD_INLINE c_v64 c_v64_shr_s8(c_v64 a, unsigned int n) {
  c_v64 t;
  int c;
789
  if (SIMD_CHECK && n > 7) {
790 791 792 793 794 795 796 797 798 799
    fprintf(stderr, "Error: Undefined s8 shift right %d\n", n);
    abort();
  }
  for (c = 0; c < 8; c++) t.s8[c] = a.s8[c] >> n;
  return t;
}

SIMD_INLINE c_v64 c_v64_shl_16(c_v64 a, unsigned int n) {
  c_v64 t;
  int c;
800
  if (SIMD_CHECK && n > 15) {
801 802 803 804 805 806 807 808 809 810
    fprintf(stderr, "Error: Undefined u16 shift left %d\n", n);
    abort();
  }
  for (c = 0; c < 4; c++) t.u16[c] = a.u16[c] << n;
  return t;
}

SIMD_INLINE c_v64 c_v64_shr_u16(c_v64 a, unsigned int n) {
  c_v64 t;
  int c;
811
  if (SIMD_CHECK && n > 15) {
812 813 814 815 816 817 818 819 820 821
    fprintf(stderr, "Error: Undefined u16 shift right %d\n", n);
    abort();
  }
  for (c = 0; c < 4; c++) t.u16[c] = a.u16[c] >> n;
  return t;
}

SIMD_INLINE c_v64 c_v64_shr_s16(c_v64 a, unsigned int n) {
  c_v64 t;
  int c;
822
  if (SIMD_CHECK && n > 15) {
823 824 825 826 827 828 829 830 831
    fprintf(stderr, "Error: undefined s16 shift right %d\n", n);
    abort();
  }
  for (c = 0; c < 4; c++) t.s16[c] = a.s16[c] >> n;
  return t;
}

SIMD_INLINE c_v64 c_v64_shl_32(c_v64 a, unsigned int n) {
  c_v64 t;
832
  if (SIMD_CHECK && n > 31) {
833 834 835 836 837 838 839 840 841 842
    fprintf(stderr, "Error: undefined u32 shift left %d\n", n);
    abort();
  }
  t.u32[1] = a.u32[1] << n;
  t.u32[0] = a.u32[0] << n;
  return t;
}

SIMD_INLINE c_v64 c_v64_shr_u32(c_v64 a, unsigned int n) {
  c_v64 t;
843
  if (SIMD_CHECK && n > 31) {
844 845 846 847 848 849 850 851 852 853
    fprintf(stderr, "Error: undefined u32 shift right %d\n", n);
    abort();
  }
  t.u32[1] = a.u32[1] >> n;
  t.u32[0] = a.u32[0] >> n;
  return t;
}

SIMD_INLINE c_v64 c_v64_shr_s32(c_v64 a, unsigned int n) {
  c_v64 t;
854
  if (SIMD_CHECK && n > 31) {
855 856 857 858 859 860 861 862
    fprintf(stderr, "Error: undefined s32 shift right %d\n", n);
    abort();
  }
  t.s32[1] = a.s32[1] >> n;
  t.s32[0] = a.s32[0] >> n;
  return t;
}

Yaowu Xu's avatar
Yaowu Xu committed
863
SIMD_INLINE c_v64 c_v64_shr_n_byte(c_v64 x, unsigned int i) {
864 865 866 867 868
  c_v64 t;
  t.u64 = x.u64 >> i * 8;
  return t;
}

Yaowu Xu's avatar
Yaowu Xu committed
869
SIMD_INLINE c_v64 c_v64_shl_n_byte(c_v64 x, unsigned int i) {
870 871 872 873 874
  c_v64 t;
  t.u64 = x.u64 << i * 8;
  return t;
}

Yaowu Xu's avatar
Yaowu Xu committed
875
SIMD_INLINE c_v64 c_v64_align(c_v64 a, c_v64 b, unsigned int c) {
876
  if (SIMD_CHECK && c > 7) {
877 878 879 880 881 882
    fprintf(stderr, "Error: undefined alignment %d\n", c);
    abort();
  }
  return c ? c_v64_or(c_v64_shr_n_byte(b, c), c_v64_shl_n_byte(a, 8 - c)) : b;
}

Yaowu Xu's avatar
Yaowu Xu committed
883
SIMD_INLINE c_v64 c_v64_shl_n_8(c_v64 a, unsigned int c) {
884 885 886
  return c_v64_shl_8(a, c);
}

Yaowu Xu's avatar
Yaowu Xu committed
887
SIMD_INLINE c_v64 c_v64_shr_n_u8(c_v64 a, unsigned int c) {
888 889 890
  return c_v64_shr_u8(a, c);
}

Yaowu Xu's avatar
Yaowu Xu committed
891
SIMD_INLINE c_v64 c_v64_shr_n_s8(c_v64 a, unsigned int c) {
892 893 894
  return c_v64_shr_s8(a, c);
}

Yaowu Xu's avatar
Yaowu Xu committed
895
SIMD_INLINE c_v64 c_v64_shl_n_16(c_v64 a, unsigned int c) {
896 897 898
  return c_v64_shl_16(a, c);
}

Yaowu Xu's avatar
Yaowu Xu committed
899
SIMD_INLINE c_v64 c_v64_shr_n_u16(c_v64 a, unsigned int c) {
900 901 902
  return c_v64_shr_u16(a, c);
}

Yaowu Xu's avatar
Yaowu Xu committed
903
SIMD_INLINE c_v64 c_v64_shr_n_s16(c_v64 a, unsigned int c) {
904 905 906
  return c_v64_shr_s16(a, c);
}

Yaowu Xu's avatar
Yaowu Xu committed
907
SIMD_INLINE c_v64 c_v64_shl_n_32(c_v64 a, unsigned int c) {
908 909 910
  return c_v64_shl_32(a, c);
}

Yaowu Xu's avatar
Yaowu Xu committed
911
SIMD_INLINE c_v64 c_v64_shr_n_u32(c_v64 a, unsigned int c) {
912 913 914
  return c_v64_shr_u32(a, c);
}

Yaowu Xu's avatar
Yaowu Xu committed
915
SIMD_INLINE c_v64 c_v64_shr_n_s32(c_v64 a, unsigned int c) {
916 917 918 919
  return c_v64_shr_s32(a, c);
}

#endif /* _V64_INTRINSICS_C_H */