av1_inv_txfm1d.c 80.3 KB
Newer Older
Angie Chiang's avatar
Angie Chiang committed
1
/*
2
 * Copyright (c) 2016, Alliance for Open Media. All rights reserved
Angie Chiang's avatar
Angie Chiang committed
3
 *
4 5 6 7 8 9
 * 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.
Angie Chiang's avatar
Angie Chiang committed
10 11
 */

12
#include <stdlib.h>
13
#include "aom_dsp/inv_txfm.h"
Yaowu Xu's avatar
Yaowu Xu committed
14
#include "av1/common/av1_inv_txfm1d.h"
Angie Chiang's avatar
Angie Chiang committed
15
#if CONFIG_COEFFICIENT_RANGE_CHECKING
16 17 18

void range_check_func(int32_t stage, const int32_t *input, const int32_t *buf,
                      int32_t size, int8_t bit) {
19 20
  const int64_t maxValue = (1LL << (bit - 1)) - 1;
  const int64_t minValue = -(1LL << (bit - 1));
21

22 23
  int in_range = 1;

24 25
  for (int i = 0; i < size; ++i) {
    if (buf[i] < minValue || buf[i] > maxValue) {
26 27 28 29 30 31
      in_range = 0;
    }
  }

  if (!in_range) {
    fprintf(stderr, "Error: coeffs contain out-of-range values\n");
32
    fprintf(stderr, "size: %d\n", size);
33 34 35 36 37
    fprintf(stderr, "stage: %d\n", stage);
    fprintf(stderr, "allowed range: [%" PRId64 ";%" PRId64 "]\n", minValue,
            maxValue);

    fprintf(stderr, "coeffs: ");
38

39 40 41 42
    fprintf(stderr, "[");
    for (int j = 0; j < size; j++) {
      if (j > 0) fprintf(stderr, ", ");
      fprintf(stderr, "%d", input[j]);
43
    }
44 45 46 47 48 49 50 51 52 53
    fprintf(stderr, "]\n");

    fprintf(stderr, "   buf: ");

    fprintf(stderr, "[");
    for (int j = 0; j < size; j++) {
      if (j > 0) fprintf(stderr, ", ");
      fprintf(stderr, "%d", buf[j]);
    }
    fprintf(stderr, "]\n\n");
Angie Chiang's avatar
Angie Chiang committed
54
  }
55 56

  assert(in_range);
57 58 59 60
}

#define range_check(stage, input, buf, size, bit) \
  range_check_func(stage, input, buf, size, bit)
Angie Chiang's avatar
Angie Chiang committed
61 62
#else
#define range_check(stage, input, buf, size, bit) \
63 64 65 66 67 68 69
  {                                               \
    (void)stage;                                  \
    (void)input;                                  \
    (void)buf;                                    \
    (void)size;                                   \
    (void)bit;                                    \
  }
Angie Chiang's avatar
Angie Chiang committed
70 71
#endif

72
// TODO(angiebird): Make 1-d txfm functions static
Yaowu Xu's avatar
Yaowu Xu committed
73 74
void av1_idct4_new(const int32_t *input, int32_t *output, const int8_t *cos_bit,
                   const int8_t *stage_range) {
Angie Chiang's avatar
Angie Chiang committed
75 76 77 78 79 80 81 82 83 84 85 86
  const int32_t size = 4;
  const int32_t *cospi;

  int32_t stage = 0;
  int32_t *bf0, *bf1;
  int32_t step[4];

  // stage 0;
  range_check(stage, input, input, size, stage_range[stage]);

  // stage 1;
  stage++;
87
  assert(output != input);
Angie Chiang's avatar
Angie Chiang committed
88 89 90 91 92 93 94 95 96
  bf1 = output;
  bf1[0] = input[0];
  bf1[1] = input[2];
  bf1[2] = input[1];
  bf1[3] = input[3];
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 2
  stage++;
97
  cospi = cospi_arr(cos_bit[stage]);
Angie Chiang's avatar
Angie Chiang committed
98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115
  bf0 = output;
  bf1 = step;
  bf1[0] = half_btf(cospi[32], bf0[0], cospi[32], bf0[1], cos_bit[stage]);
  bf1[1] = half_btf(cospi[32], bf0[0], -cospi[32], bf0[1], cos_bit[stage]);
  bf1[2] = half_btf(cospi[48], bf0[2], -cospi[16], bf0[3], cos_bit[stage]);
  bf1[3] = half_btf(cospi[16], bf0[2], cospi[48], bf0[3], cos_bit[stage]);
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 3
  stage++;
  bf0 = step;
  bf1 = output;
  bf1[0] = bf0[0] + bf0[3];
  bf1[1] = bf0[1] + bf0[2];
  bf1[2] = bf0[1] - bf0[2];
  bf1[3] = bf0[0] - bf0[3];
  range_check(stage, input, bf1, size, stage_range[stage]);
}
Angie Chiang's avatar
Angie Chiang committed
116

Yaowu Xu's avatar
Yaowu Xu committed
117 118
void av1_idct8_new(const int32_t *input, int32_t *output, const int8_t *cos_bit,
                   const int8_t *stage_range) {
Angie Chiang's avatar
Angie Chiang committed
119 120 121 122 123 124 125 126 127 128 129 130
  const int32_t size = 8;
  const int32_t *cospi;

  int32_t stage = 0;
  int32_t *bf0, *bf1;
  int32_t step[8];

  // stage 0;
  range_check(stage, input, input, size, stage_range[stage]);

  // stage 1;
  stage++;
131
  assert(output != input);
Angie Chiang's avatar
Angie Chiang committed
132 133 134 135 136 137 138 139 140 141 142 143 144
  bf1 = output;
  bf1[0] = input[0];
  bf1[1] = input[4];
  bf1[2] = input[2];
  bf1[3] = input[6];
  bf1[4] = input[1];
  bf1[5] = input[5];
  bf1[6] = input[3];
  bf1[7] = input[7];
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 2
  stage++;
145
  cospi = cospi_arr(cos_bit[stage]);
Angie Chiang's avatar
Angie Chiang committed
146 147 148 149 150 151 152 153 154 155 156 157 158 159
  bf0 = output;
  bf1 = step;
  bf1[0] = bf0[0];
  bf1[1] = bf0[1];
  bf1[2] = bf0[2];
  bf1[3] = bf0[3];
  bf1[4] = half_btf(cospi[56], bf0[4], -cospi[8], bf0[7], cos_bit[stage]);
  bf1[5] = half_btf(cospi[24], bf0[5], -cospi[40], bf0[6], cos_bit[stage]);
  bf1[6] = half_btf(cospi[40], bf0[5], cospi[24], bf0[6], cos_bit[stage]);
  bf1[7] = half_btf(cospi[8], bf0[4], cospi[56], bf0[7], cos_bit[stage]);
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 3
  stage++;
160
  cospi = cospi_arr(cos_bit[stage]);
Angie Chiang's avatar
Angie Chiang committed
161 162 163 164 165 166 167 168 169 170 171 172 173 174
  bf0 = step;
  bf1 = output;
  bf1[0] = half_btf(cospi[32], bf0[0], cospi[32], bf0[1], cos_bit[stage]);
  bf1[1] = half_btf(cospi[32], bf0[0], -cospi[32], bf0[1], cos_bit[stage]);
  bf1[2] = half_btf(cospi[48], bf0[2], -cospi[16], bf0[3], cos_bit[stage]);
  bf1[3] = half_btf(cospi[16], bf0[2], cospi[48], bf0[3], cos_bit[stage]);
  bf1[4] = bf0[4] + bf0[5];
  bf1[5] = bf0[4] - bf0[5];
  bf1[6] = -bf0[6] + bf0[7];
  bf1[7] = bf0[6] + bf0[7];
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 4
  stage++;
175
  cospi = cospi_arr(cos_bit[stage]);
Angie Chiang's avatar
Angie Chiang committed
176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201
  bf0 = output;
  bf1 = step;
  bf1[0] = bf0[0] + bf0[3];
  bf1[1] = bf0[1] + bf0[2];
  bf1[2] = bf0[1] - bf0[2];
  bf1[3] = bf0[0] - bf0[3];
  bf1[4] = bf0[4];
  bf1[5] = half_btf(-cospi[32], bf0[5], cospi[32], bf0[6], cos_bit[stage]);
  bf1[6] = half_btf(cospi[32], bf0[5], cospi[32], bf0[6], cos_bit[stage]);
  bf1[7] = bf0[7];
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 5
  stage++;
  bf0 = step;
  bf1 = output;
  bf1[0] = bf0[0] + bf0[7];
  bf1[1] = bf0[1] + bf0[6];
  bf1[2] = bf0[2] + bf0[5];
  bf1[3] = bf0[3] + bf0[4];
  bf1[4] = bf0[3] - bf0[4];
  bf1[5] = bf0[2] - bf0[5];
  bf1[6] = bf0[1] - bf0[6];
  bf1[7] = bf0[0] - bf0[7];
  range_check(stage, input, bf1, size, stage_range[stage]);
}
Angie Chiang's avatar
Angie Chiang committed
202

Yaowu Xu's avatar
Yaowu Xu committed
203 204
void av1_idct16_new(const int32_t *input, int32_t *output,
                    const int8_t *cos_bit, const int8_t *stage_range) {
Angie Chiang's avatar
Angie Chiang committed
205 206 207 208 209 210 211 212 213 214 215 216
  const int32_t size = 16;
  const int32_t *cospi;

  int32_t stage = 0;
  int32_t *bf0, *bf1;
  int32_t step[16];

  // stage 0;
  range_check(stage, input, input, size, stage_range[stage]);

  // stage 1;
  stage++;
217
  assert(output != input);
Angie Chiang's avatar
Angie Chiang committed
218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238
  bf1 = output;
  bf1[0] = input[0];
  bf1[1] = input[8];
  bf1[2] = input[4];
  bf1[3] = input[12];
  bf1[4] = input[2];
  bf1[5] = input[10];
  bf1[6] = input[6];
  bf1[7] = input[14];
  bf1[8] = input[1];
  bf1[9] = input[9];
  bf1[10] = input[5];
  bf1[11] = input[13];
  bf1[12] = input[3];
  bf1[13] = input[11];
  bf1[14] = input[7];
  bf1[15] = input[15];
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 2
  stage++;
239
  cospi = cospi_arr(cos_bit[stage]);
Angie Chiang's avatar
Angie Chiang committed
240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261
  bf0 = output;
  bf1 = step;
  bf1[0] = bf0[0];
  bf1[1] = bf0[1];
  bf1[2] = bf0[2];
  bf1[3] = bf0[3];
  bf1[4] = bf0[4];
  bf1[5] = bf0[5];
  bf1[6] = bf0[6];
  bf1[7] = bf0[7];
  bf1[8] = half_btf(cospi[60], bf0[8], -cospi[4], bf0[15], cos_bit[stage]);
  bf1[9] = half_btf(cospi[28], bf0[9], -cospi[36], bf0[14], cos_bit[stage]);
  bf1[10] = half_btf(cospi[44], bf0[10], -cospi[20], bf0[13], cos_bit[stage]);
  bf1[11] = half_btf(cospi[12], bf0[11], -cospi[52], bf0[12], cos_bit[stage]);
  bf1[12] = half_btf(cospi[52], bf0[11], cospi[12], bf0[12], cos_bit[stage]);
  bf1[13] = half_btf(cospi[20], bf0[10], cospi[44], bf0[13], cos_bit[stage]);
  bf1[14] = half_btf(cospi[36], bf0[9], cospi[28], bf0[14], cos_bit[stage]);
  bf1[15] = half_btf(cospi[4], bf0[8], cospi[60], bf0[15], cos_bit[stage]);
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 3
  stage++;
262
  cospi = cospi_arr(cos_bit[stage]);
Angie Chiang's avatar
Angie Chiang committed
263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284
  bf0 = step;
  bf1 = output;
  bf1[0] = bf0[0];
  bf1[1] = bf0[1];
  bf1[2] = bf0[2];
  bf1[3] = bf0[3];
  bf1[4] = half_btf(cospi[56], bf0[4], -cospi[8], bf0[7], cos_bit[stage]);
  bf1[5] = half_btf(cospi[24], bf0[5], -cospi[40], bf0[6], cos_bit[stage]);
  bf1[6] = half_btf(cospi[40], bf0[5], cospi[24], bf0[6], cos_bit[stage]);
  bf1[7] = half_btf(cospi[8], bf0[4], cospi[56], bf0[7], cos_bit[stage]);
  bf1[8] = bf0[8] + bf0[9];
  bf1[9] = bf0[8] - bf0[9];
  bf1[10] = -bf0[10] + bf0[11];
  bf1[11] = bf0[10] + bf0[11];
  bf1[12] = bf0[12] + bf0[13];
  bf1[13] = bf0[12] - bf0[13];
  bf1[14] = -bf0[14] + bf0[15];
  bf1[15] = bf0[14] + bf0[15];
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 4
  stage++;
285
  cospi = cospi_arr(cos_bit[stage]);
Angie Chiang's avatar
Angie Chiang committed
286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307
  bf0 = output;
  bf1 = step;
  bf1[0] = half_btf(cospi[32], bf0[0], cospi[32], bf0[1], cos_bit[stage]);
  bf1[1] = half_btf(cospi[32], bf0[0], -cospi[32], bf0[1], cos_bit[stage]);
  bf1[2] = half_btf(cospi[48], bf0[2], -cospi[16], bf0[3], cos_bit[stage]);
  bf1[3] = half_btf(cospi[16], bf0[2], cospi[48], bf0[3], cos_bit[stage]);
  bf1[4] = bf0[4] + bf0[5];
  bf1[5] = bf0[4] - bf0[5];
  bf1[6] = -bf0[6] + bf0[7];
  bf1[7] = bf0[6] + bf0[7];
  bf1[8] = bf0[8];
  bf1[9] = half_btf(-cospi[16], bf0[9], cospi[48], bf0[14], cos_bit[stage]);
  bf1[10] = half_btf(-cospi[48], bf0[10], -cospi[16], bf0[13], cos_bit[stage]);
  bf1[11] = bf0[11];
  bf1[12] = bf0[12];
  bf1[13] = half_btf(-cospi[16], bf0[10], cospi[48], bf0[13], cos_bit[stage]);
  bf1[14] = half_btf(cospi[48], bf0[9], cospi[16], bf0[14], cos_bit[stage]);
  bf1[15] = bf0[15];
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 5
  stage++;
308
  cospi = cospi_arr(cos_bit[stage]);
Angie Chiang's avatar
Angie Chiang committed
309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330
  bf0 = step;
  bf1 = output;
  bf1[0] = bf0[0] + bf0[3];
  bf1[1] = bf0[1] + bf0[2];
  bf1[2] = bf0[1] - bf0[2];
  bf1[3] = bf0[0] - bf0[3];
  bf1[4] = bf0[4];
  bf1[5] = half_btf(-cospi[32], bf0[5], cospi[32], bf0[6], cos_bit[stage]);
  bf1[6] = half_btf(cospi[32], bf0[5], cospi[32], bf0[6], cos_bit[stage]);
  bf1[7] = bf0[7];
  bf1[8] = bf0[8] + bf0[11];
  bf1[9] = bf0[9] + bf0[10];
  bf1[10] = bf0[9] - bf0[10];
  bf1[11] = bf0[8] - bf0[11];
  bf1[12] = -bf0[12] + bf0[15];
  bf1[13] = -bf0[13] + bf0[14];
  bf1[14] = bf0[13] + bf0[14];
  bf1[15] = bf0[12] + bf0[15];
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 6
  stage++;
331
  cospi = cospi_arr(cos_bit[stage]);
Angie Chiang's avatar
Angie Chiang committed
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
  bf0 = output;
  bf1 = step;
  bf1[0] = bf0[0] + bf0[7];
  bf1[1] = bf0[1] + bf0[6];
  bf1[2] = bf0[2] + bf0[5];
  bf1[3] = bf0[3] + bf0[4];
  bf1[4] = bf0[3] - bf0[4];
  bf1[5] = bf0[2] - bf0[5];
  bf1[6] = bf0[1] - bf0[6];
  bf1[7] = bf0[0] - bf0[7];
  bf1[8] = bf0[8];
  bf1[9] = bf0[9];
  bf1[10] = half_btf(-cospi[32], bf0[10], cospi[32], bf0[13], cos_bit[stage]);
  bf1[11] = half_btf(-cospi[32], bf0[11], cospi[32], bf0[12], cos_bit[stage]);
  bf1[12] = half_btf(cospi[32], bf0[11], cospi[32], bf0[12], cos_bit[stage]);
  bf1[13] = half_btf(cospi[32], bf0[10], cospi[32], bf0[13], cos_bit[stage]);
  bf1[14] = bf0[14];
  bf1[15] = bf0[15];
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 7
  stage++;
  bf0 = step;
  bf1 = output;
  bf1[0] = bf0[0] + bf0[15];
  bf1[1] = bf0[1] + bf0[14];
  bf1[2] = bf0[2] + bf0[13];
  bf1[3] = bf0[3] + bf0[12];
  bf1[4] = bf0[4] + bf0[11];
  bf1[5] = bf0[5] + bf0[10];
  bf1[6] = bf0[6] + bf0[9];
  bf1[7] = bf0[7] + bf0[8];
  bf1[8] = bf0[7] - bf0[8];
  bf1[9] = bf0[6] - bf0[9];
  bf1[10] = bf0[5] - bf0[10];
  bf1[11] = bf0[4] - bf0[11];
  bf1[12] = bf0[3] - bf0[12];
  bf1[13] = bf0[2] - bf0[13];
  bf1[14] = bf0[1] - bf0[14];
  bf1[15] = bf0[0] - bf0[15];
  range_check(stage, input, bf1, size, stage_range[stage]);
}
Angie Chiang's avatar
Angie Chiang committed
374

Yaowu Xu's avatar
Yaowu Xu committed
375 376
void av1_idct32_new(const int32_t *input, int32_t *output,
                    const int8_t *cos_bit, const int8_t *stage_range) {
Angie Chiang's avatar
Angie Chiang committed
377 378 379 380 381 382 383 384 385 386 387 388
  const int32_t size = 32;
  const int32_t *cospi;

  int32_t stage = 0;
  int32_t *bf0, *bf1;
  int32_t step[32];

  // stage 0;
  range_check(stage, input, input, size, stage_range[stage]);

  // stage 1;
  stage++;
389
  assert(output != input);
Angie Chiang's avatar
Angie Chiang committed
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
  bf1 = output;
  bf1[0] = input[0];
  bf1[1] = input[16];
  bf1[2] = input[8];
  bf1[3] = input[24];
  bf1[4] = input[4];
  bf1[5] = input[20];
  bf1[6] = input[12];
  bf1[7] = input[28];
  bf1[8] = input[2];
  bf1[9] = input[18];
  bf1[10] = input[10];
  bf1[11] = input[26];
  bf1[12] = input[6];
  bf1[13] = input[22];
  bf1[14] = input[14];
  bf1[15] = input[30];
  bf1[16] = input[1];
  bf1[17] = input[17];
  bf1[18] = input[9];
  bf1[19] = input[25];
  bf1[20] = input[5];
  bf1[21] = input[21];
  bf1[22] = input[13];
  bf1[23] = input[29];
  bf1[24] = input[3];
  bf1[25] = input[19];
  bf1[26] = input[11];
  bf1[27] = input[27];
  bf1[28] = input[7];
  bf1[29] = input[23];
  bf1[30] = input[15];
  bf1[31] = input[31];
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 2
  stage++;
427
  cospi = cospi_arr(cos_bit[stage]);
Angie Chiang's avatar
Angie Chiang committed
428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465
  bf0 = output;
  bf1 = step;
  bf1[0] = bf0[0];
  bf1[1] = bf0[1];
  bf1[2] = bf0[2];
  bf1[3] = bf0[3];
  bf1[4] = bf0[4];
  bf1[5] = bf0[5];
  bf1[6] = bf0[6];
  bf1[7] = bf0[7];
  bf1[8] = bf0[8];
  bf1[9] = bf0[9];
  bf1[10] = bf0[10];
  bf1[11] = bf0[11];
  bf1[12] = bf0[12];
  bf1[13] = bf0[13];
  bf1[14] = bf0[14];
  bf1[15] = bf0[15];
  bf1[16] = half_btf(cospi[62], bf0[16], -cospi[2], bf0[31], cos_bit[stage]);
  bf1[17] = half_btf(cospi[30], bf0[17], -cospi[34], bf0[30], cos_bit[stage]);
  bf1[18] = half_btf(cospi[46], bf0[18], -cospi[18], bf0[29], cos_bit[stage]);
  bf1[19] = half_btf(cospi[14], bf0[19], -cospi[50], bf0[28], cos_bit[stage]);
  bf1[20] = half_btf(cospi[54], bf0[20], -cospi[10], bf0[27], cos_bit[stage]);
  bf1[21] = half_btf(cospi[22], bf0[21], -cospi[42], bf0[26], cos_bit[stage]);
  bf1[22] = half_btf(cospi[38], bf0[22], -cospi[26], bf0[25], cos_bit[stage]);
  bf1[23] = half_btf(cospi[6], bf0[23], -cospi[58], bf0[24], cos_bit[stage]);
  bf1[24] = half_btf(cospi[58], bf0[23], cospi[6], bf0[24], cos_bit[stage]);
  bf1[25] = half_btf(cospi[26], bf0[22], cospi[38], bf0[25], cos_bit[stage]);
  bf1[26] = half_btf(cospi[42], bf0[21], cospi[22], bf0[26], cos_bit[stage]);
  bf1[27] = half_btf(cospi[10], bf0[20], cospi[54], bf0[27], cos_bit[stage]);
  bf1[28] = half_btf(cospi[50], bf0[19], cospi[14], bf0[28], cos_bit[stage]);
  bf1[29] = half_btf(cospi[18], bf0[18], cospi[46], bf0[29], cos_bit[stage]);
  bf1[30] = half_btf(cospi[34], bf0[17], cospi[30], bf0[30], cos_bit[stage]);
  bf1[31] = half_btf(cospi[2], bf0[16], cospi[62], bf0[31], cos_bit[stage]);
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 3
  stage++;
466
  cospi = cospi_arr(cos_bit[stage]);
Angie Chiang's avatar
Angie Chiang committed
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
  bf0 = step;
  bf1 = output;
  bf1[0] = bf0[0];
  bf1[1] = bf0[1];
  bf1[2] = bf0[2];
  bf1[3] = bf0[3];
  bf1[4] = bf0[4];
  bf1[5] = bf0[5];
  bf1[6] = bf0[6];
  bf1[7] = bf0[7];
  bf1[8] = half_btf(cospi[60], bf0[8], -cospi[4], bf0[15], cos_bit[stage]);
  bf1[9] = half_btf(cospi[28], bf0[9], -cospi[36], bf0[14], cos_bit[stage]);
  bf1[10] = half_btf(cospi[44], bf0[10], -cospi[20], bf0[13], cos_bit[stage]);
  bf1[11] = half_btf(cospi[12], bf0[11], -cospi[52], bf0[12], cos_bit[stage]);
  bf1[12] = half_btf(cospi[52], bf0[11], cospi[12], bf0[12], cos_bit[stage]);
  bf1[13] = half_btf(cospi[20], bf0[10], cospi[44], bf0[13], cos_bit[stage]);
  bf1[14] = half_btf(cospi[36], bf0[9], cospi[28], bf0[14], cos_bit[stage]);
  bf1[15] = half_btf(cospi[4], bf0[8], cospi[60], bf0[15], cos_bit[stage]);
  bf1[16] = bf0[16] + bf0[17];
  bf1[17] = bf0[16] - bf0[17];
  bf1[18] = -bf0[18] + bf0[19];
  bf1[19] = bf0[18] + bf0[19];
  bf1[20] = bf0[20] + bf0[21];
  bf1[21] = bf0[20] - bf0[21];
  bf1[22] = -bf0[22] + bf0[23];
  bf1[23] = bf0[22] + bf0[23];
  bf1[24] = bf0[24] + bf0[25];
  bf1[25] = bf0[24] - bf0[25];
  bf1[26] = -bf0[26] + bf0[27];
  bf1[27] = bf0[26] + bf0[27];
  bf1[28] = bf0[28] + bf0[29];
  bf1[29] = bf0[28] - bf0[29];
  bf1[30] = -bf0[30] + bf0[31];
  bf1[31] = bf0[30] + bf0[31];
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 4
  stage++;
505
  cospi = cospi_arr(cos_bit[stage]);
Angie Chiang's avatar
Angie Chiang committed
506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543
  bf0 = output;
  bf1 = step;
  bf1[0] = bf0[0];
  bf1[1] = bf0[1];
  bf1[2] = bf0[2];
  bf1[3] = bf0[3];
  bf1[4] = half_btf(cospi[56], bf0[4], -cospi[8], bf0[7], cos_bit[stage]);
  bf1[5] = half_btf(cospi[24], bf0[5], -cospi[40], bf0[6], cos_bit[stage]);
  bf1[6] = half_btf(cospi[40], bf0[5], cospi[24], bf0[6], cos_bit[stage]);
  bf1[7] = half_btf(cospi[8], bf0[4], cospi[56], bf0[7], cos_bit[stage]);
  bf1[8] = bf0[8] + bf0[9];
  bf1[9] = bf0[8] - bf0[9];
  bf1[10] = -bf0[10] + bf0[11];
  bf1[11] = bf0[10] + bf0[11];
  bf1[12] = bf0[12] + bf0[13];
  bf1[13] = bf0[12] - bf0[13];
  bf1[14] = -bf0[14] + bf0[15];
  bf1[15] = bf0[14] + bf0[15];
  bf1[16] = bf0[16];
  bf1[17] = half_btf(-cospi[8], bf0[17], cospi[56], bf0[30], cos_bit[stage]);
  bf1[18] = half_btf(-cospi[56], bf0[18], -cospi[8], bf0[29], cos_bit[stage]);
  bf1[19] = bf0[19];
  bf1[20] = bf0[20];
  bf1[21] = half_btf(-cospi[40], bf0[21], cospi[24], bf0[26], cos_bit[stage]);
  bf1[22] = half_btf(-cospi[24], bf0[22], -cospi[40], bf0[25], cos_bit[stage]);
  bf1[23] = bf0[23];
  bf1[24] = bf0[24];
  bf1[25] = half_btf(-cospi[40], bf0[22], cospi[24], bf0[25], cos_bit[stage]);
  bf1[26] = half_btf(cospi[24], bf0[21], cospi[40], bf0[26], cos_bit[stage]);
  bf1[27] = bf0[27];
  bf1[28] = bf0[28];
  bf1[29] = half_btf(-cospi[8], bf0[18], cospi[56], bf0[29], cos_bit[stage]);
  bf1[30] = half_btf(cospi[56], bf0[17], cospi[8], bf0[30], cos_bit[stage]);
  bf1[31] = bf0[31];
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 5
  stage++;
544
  cospi = cospi_arr(cos_bit[stage]);
Angie Chiang's avatar
Angie Chiang committed
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
  bf0 = step;
  bf1 = output;
  bf1[0] = half_btf(cospi[32], bf0[0], cospi[32], bf0[1], cos_bit[stage]);
  bf1[1] = half_btf(cospi[32], bf0[0], -cospi[32], bf0[1], cos_bit[stage]);
  bf1[2] = half_btf(cospi[48], bf0[2], -cospi[16], bf0[3], cos_bit[stage]);
  bf1[3] = half_btf(cospi[16], bf0[2], cospi[48], bf0[3], cos_bit[stage]);
  bf1[4] = bf0[4] + bf0[5];
  bf1[5] = bf0[4] - bf0[5];
  bf1[6] = -bf0[6] + bf0[7];
  bf1[7] = bf0[6] + bf0[7];
  bf1[8] = bf0[8];
  bf1[9] = half_btf(-cospi[16], bf0[9], cospi[48], bf0[14], cos_bit[stage]);
  bf1[10] = half_btf(-cospi[48], bf0[10], -cospi[16], bf0[13], cos_bit[stage]);
  bf1[11] = bf0[11];
  bf1[12] = bf0[12];
  bf1[13] = half_btf(-cospi[16], bf0[10], cospi[48], bf0[13], cos_bit[stage]);
  bf1[14] = half_btf(cospi[48], bf0[9], cospi[16], bf0[14], cos_bit[stage]);
  bf1[15] = bf0[15];
  bf1[16] = bf0[16] + bf0[19];
  bf1[17] = bf0[17] + bf0[18];
  bf1[18] = bf0[17] - bf0[18];
  bf1[19] = bf0[16] - bf0[19];
  bf1[20] = -bf0[20] + bf0[23];
  bf1[21] = -bf0[21] + bf0[22];
  bf1[22] = bf0[21] + bf0[22];
  bf1[23] = bf0[20] + bf0[23];
  bf1[24] = bf0[24] + bf0[27];
  bf1[25] = bf0[25] + bf0[26];
  bf1[26] = bf0[25] - bf0[26];
  bf1[27] = bf0[24] - bf0[27];
  bf1[28] = -bf0[28] + bf0[31];
  bf1[29] = -bf0[29] + bf0[30];
  bf1[30] = bf0[29] + bf0[30];
  bf1[31] = bf0[28] + bf0[31];
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 6
  stage++;
583
  cospi = cospi_arr(cos_bit[stage]);
Angie Chiang's avatar
Angie Chiang committed
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
  bf0 = output;
  bf1 = step;
  bf1[0] = bf0[0] + bf0[3];
  bf1[1] = bf0[1] + bf0[2];
  bf1[2] = bf0[1] - bf0[2];
  bf1[3] = bf0[0] - bf0[3];
  bf1[4] = bf0[4];
  bf1[5] = half_btf(-cospi[32], bf0[5], cospi[32], bf0[6], cos_bit[stage]);
  bf1[6] = half_btf(cospi[32], bf0[5], cospi[32], bf0[6], cos_bit[stage]);
  bf1[7] = bf0[7];
  bf1[8] = bf0[8] + bf0[11];
  bf1[9] = bf0[9] + bf0[10];
  bf1[10] = bf0[9] - bf0[10];
  bf1[11] = bf0[8] - bf0[11];
  bf1[12] = -bf0[12] + bf0[15];
  bf1[13] = -bf0[13] + bf0[14];
  bf1[14] = bf0[13] + bf0[14];
  bf1[15] = bf0[12] + bf0[15];
  bf1[16] = bf0[16];
  bf1[17] = bf0[17];
  bf1[18] = half_btf(-cospi[16], bf0[18], cospi[48], bf0[29], cos_bit[stage]);
  bf1[19] = half_btf(-cospi[16], bf0[19], cospi[48], bf0[28], cos_bit[stage]);
  bf1[20] = half_btf(-cospi[48], bf0[20], -cospi[16], bf0[27], cos_bit[stage]);
  bf1[21] = half_btf(-cospi[48], bf0[21], -cospi[16], bf0[26], cos_bit[stage]);
  bf1[22] = bf0[22];
  bf1[23] = bf0[23];
  bf1[24] = bf0[24];
  bf1[25] = bf0[25];
  bf1[26] = half_btf(-cospi[16], bf0[21], cospi[48], bf0[26], cos_bit[stage]);
  bf1[27] = half_btf(-cospi[16], bf0[20], cospi[48], bf0[27], cos_bit[stage]);
  bf1[28] = half_btf(cospi[48], bf0[19], cospi[16], bf0[28], cos_bit[stage]);
  bf1[29] = half_btf(cospi[48], bf0[18], cospi[16], bf0[29], cos_bit[stage]);
  bf1[30] = bf0[30];
  bf1[31] = bf0[31];
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 7
  stage++;
622
  cospi = cospi_arr(cos_bit[stage]);
Angie Chiang's avatar
Angie Chiang committed
623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660
  bf0 = step;
  bf1 = output;
  bf1[0] = bf0[0] + bf0[7];
  bf1[1] = bf0[1] + bf0[6];
  bf1[2] = bf0[2] + bf0[5];
  bf1[3] = bf0[3] + bf0[4];
  bf1[4] = bf0[3] - bf0[4];
  bf1[5] = bf0[2] - bf0[5];
  bf1[6] = bf0[1] - bf0[6];
  bf1[7] = bf0[0] - bf0[7];
  bf1[8] = bf0[8];
  bf1[9] = bf0[9];
  bf1[10] = half_btf(-cospi[32], bf0[10], cospi[32], bf0[13], cos_bit[stage]);
  bf1[11] = half_btf(-cospi[32], bf0[11], cospi[32], bf0[12], cos_bit[stage]);
  bf1[12] = half_btf(cospi[32], bf0[11], cospi[32], bf0[12], cos_bit[stage]);
  bf1[13] = half_btf(cospi[32], bf0[10], cospi[32], bf0[13], cos_bit[stage]);
  bf1[14] = bf0[14];
  bf1[15] = bf0[15];
  bf1[16] = bf0[16] + bf0[23];
  bf1[17] = bf0[17] + bf0[22];
  bf1[18] = bf0[18] + bf0[21];
  bf1[19] = bf0[19] + bf0[20];
  bf1[20] = bf0[19] - bf0[20];
  bf1[21] = bf0[18] - bf0[21];
  bf1[22] = bf0[17] - bf0[22];
  bf1[23] = bf0[16] - bf0[23];
  bf1[24] = -bf0[24] + bf0[31];
  bf1[25] = -bf0[25] + bf0[30];
  bf1[26] = -bf0[26] + bf0[29];
  bf1[27] = -bf0[27] + bf0[28];
  bf1[28] = bf0[27] + bf0[28];
  bf1[29] = bf0[26] + bf0[29];
  bf1[30] = bf0[25] + bf0[30];
  bf1[31] = bf0[24] + bf0[31];
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 8
  stage++;
661
  cospi = cospi_arr(cos_bit[stage]);
Angie Chiang's avatar
Angie Chiang committed
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
  bf0 = output;
  bf1 = step;
  bf1[0] = bf0[0] + bf0[15];
  bf1[1] = bf0[1] + bf0[14];
  bf1[2] = bf0[2] + bf0[13];
  bf1[3] = bf0[3] + bf0[12];
  bf1[4] = bf0[4] + bf0[11];
  bf1[5] = bf0[5] + bf0[10];
  bf1[6] = bf0[6] + bf0[9];
  bf1[7] = bf0[7] + bf0[8];
  bf1[8] = bf0[7] - bf0[8];
  bf1[9] = bf0[6] - bf0[9];
  bf1[10] = bf0[5] - bf0[10];
  bf1[11] = bf0[4] - bf0[11];
  bf1[12] = bf0[3] - bf0[12];
  bf1[13] = bf0[2] - bf0[13];
  bf1[14] = bf0[1] - bf0[14];
  bf1[15] = bf0[0] - bf0[15];
  bf1[16] = bf0[16];
  bf1[17] = bf0[17];
  bf1[18] = bf0[18];
  bf1[19] = bf0[19];
  bf1[20] = half_btf(-cospi[32], bf0[20], cospi[32], bf0[27], cos_bit[stage]);
  bf1[21] = half_btf(-cospi[32], bf0[21], cospi[32], bf0[26], cos_bit[stage]);
  bf1[22] = half_btf(-cospi[32], bf0[22], cospi[32], bf0[25], cos_bit[stage]);
  bf1[23] = half_btf(-cospi[32], bf0[23], cospi[32], bf0[24], cos_bit[stage]);
  bf1[24] = half_btf(cospi[32], bf0[23], cospi[32], bf0[24], cos_bit[stage]);
  bf1[25] = half_btf(cospi[32], bf0[22], cospi[32], bf0[25], cos_bit[stage]);
  bf1[26] = half_btf(cospi[32], bf0[21], cospi[32], bf0[26], cos_bit[stage]);
  bf1[27] = half_btf(cospi[32], bf0[20], cospi[32], bf0[27], cos_bit[stage]);
  bf1[28] = bf0[28];
  bf1[29] = bf0[29];
  bf1[30] = bf0[30];
  bf1[31] = bf0[31];
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 9
  stage++;
  bf0 = step;
  bf1 = output;
  bf1[0] = bf0[0] + bf0[31];
  bf1[1] = bf0[1] + bf0[30];
  bf1[2] = bf0[2] + bf0[29];
  bf1[3] = bf0[3] + bf0[28];
  bf1[4] = bf0[4] + bf0[27];
  bf1[5] = bf0[5] + bf0[26];
  bf1[6] = bf0[6] + bf0[25];
  bf1[7] = bf0[7] + bf0[24];
  bf1[8] = bf0[8] + bf0[23];
  bf1[9] = bf0[9] + bf0[22];
  bf1[10] = bf0[10] + bf0[21];
  bf1[11] = bf0[11] + bf0[20];
  bf1[12] = bf0[12] + bf0[19];
  bf1[13] = bf0[13] + bf0[18];
  bf1[14] = bf0[14] + bf0[17];
  bf1[15] = bf0[15] + bf0[16];
  bf1[16] = bf0[15] - bf0[16];
  bf1[17] = bf0[14] - bf0[17];
  bf1[18] = bf0[13] - bf0[18];
  bf1[19] = bf0[12] - bf0[19];
  bf1[20] = bf0[11] - bf0[20];
  bf1[21] = bf0[10] - bf0[21];
  bf1[22] = bf0[9] - bf0[22];
  bf1[23] = bf0[8] - bf0[23];
  bf1[24] = bf0[7] - bf0[24];
  bf1[25] = bf0[6] - bf0[25];
  bf1[26] = bf0[5] - bf0[26];
  bf1[27] = bf0[4] - bf0[27];
  bf1[28] = bf0[3] - bf0[28];
  bf1[29] = bf0[2] - bf0[29];
  bf1[30] = bf0[1] - bf0[30];
  bf1[31] = bf0[0] - bf0[31];
  range_check(stage, input, bf1, size, stage_range[stage]);
}
Angie Chiang's avatar
Angie Chiang committed
736

Yaowu Xu's avatar
Yaowu Xu committed
737 738
void av1_iadst4_new(const int32_t *input, int32_t *output,
                    const int8_t *cos_bit, const int8_t *stage_range) {
739 740 741 742 743 744 745 746 747 748 749 750 751
  (void)cos_bit;
  int bd = stage_range[0];
  int64_t s0, s1, s2, s3, s4, s5, s6, s7;

  int32_t x0 = input[0];
  int32_t x1 = input[1];
  int32_t x2 = input[2];
  int32_t x3 = input[3];

  if (!(x0 | x1 | x2 | x3)) {
    output[0] = output[1] = output[2] = output[3] = 0;
    return;
  }
Angie Chiang's avatar
Angie Chiang committed
752

753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774
  s0 = sinpi_1_9 * x0;
  s1 = sinpi_2_9 * x0;
  s2 = sinpi_3_9 * x1;
  s3 = sinpi_4_9 * x2;
  s4 = sinpi_1_9 * x2;
  s5 = sinpi_2_9 * x3;
  s6 = sinpi_4_9 * x3;
  s7 = HIGHBD_WRAPLOW(x0 - x2 + x3, bd);

  s0 = s0 + s3 + s5;
  s1 = s1 - s4 - s6;
  s3 = s2;
  s2 = sinpi_3_9 * s7;

  // 1-D transform scaling factor is sqrt(2).
  // The overall dynamic range is 14b (input) + 14b (multiplication scaling)
  // + 1b (addition) = 29b.
  // Hence the output bit depth is 15b.
  output[0] = HIGHBD_WRAPLOW(dct_const_round_shift(s0 + s3), bd);
  output[1] = HIGHBD_WRAPLOW(dct_const_round_shift(s1 + s3), bd);
  output[2] = HIGHBD_WRAPLOW(dct_const_round_shift(s2), bd);
  output[3] = HIGHBD_WRAPLOW(dct_const_round_shift(s0 + s1 - s3), bd);
Angie Chiang's avatar
Angie Chiang committed
775
}
Angie Chiang's avatar
Angie Chiang committed
776

Yaowu Xu's avatar
Yaowu Xu committed
777 778
void av1_iadst8_new(const int32_t *input, int32_t *output,
                    const int8_t *cos_bit, const int8_t *stage_range) {
Angie Chiang's avatar
Angie Chiang committed
779 780 781 782 783 784 785 786 787 788 789 790 791
  const int32_t size = 8;
  const int32_t *cospi;

  int32_t stage = 0;
  int32_t *bf0, *bf1;
  int32_t step[8];

  // stage 0;
  range_check(stage, input, input, size, stage_range[stage]);

  // stage 1;
  stage++;
  bf1 = output;
792 793 794 795 796 797 798 799
  bf1[0] = input[7];
  bf1[1] = input[0];
  bf1[2] = input[5];
  bf1[3] = input[2];
  bf1[4] = input[3];
  bf1[5] = input[4];
  bf1[6] = input[1];
  bf1[7] = input[6];
Angie Chiang's avatar
Angie Chiang committed
800 801 802 803
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 2
  stage++;
804
  cospi = cospi_arr(cos_bit[stage]);
Angie Chiang's avatar
Angie Chiang committed
805 806
  bf0 = output;
  bf1 = step;
807 808 809 810 811 812 813 814
  bf1[0] = half_btf(cospi[4], bf0[0], cospi[60], bf0[1], cos_bit[stage]);
  bf1[1] = half_btf(-cospi[4], bf0[1], cospi[60], bf0[0], cos_bit[stage]);
  bf1[2] = half_btf(cospi[20], bf0[2], cospi[44], bf0[3], cos_bit[stage]);
  bf1[3] = half_btf(-cospi[20], bf0[3], cospi[44], bf0[2], cos_bit[stage]);
  bf1[4] = half_btf(cospi[36], bf0[4], cospi[28], bf0[5], cos_bit[stage]);
  bf1[5] = half_btf(-cospi[36], bf0[5], cospi[28], bf0[4], cos_bit[stage]);
  bf1[6] = half_btf(cospi[52], bf0[6], cospi[12], bf0[7], cos_bit[stage]);
  bf1[7] = half_btf(-cospi[52], bf0[7], cospi[12], bf0[6], cos_bit[stage]);
Angie Chiang's avatar
Angie Chiang committed
815 816 817 818 819 820
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 3
  stage++;
  bf0 = step;
  bf1 = output;
821 822 823 824 825 826 827 828
  bf1[0] = bf0[0] + bf0[4];
  bf1[1] = bf0[1] + bf0[5];
  bf1[2] = bf0[2] + bf0[6];
  bf1[3] = bf0[3] + bf0[7];
  bf1[4] = -bf0[4] + bf0[0];
  bf1[5] = -bf0[5] + bf0[1];
  bf1[6] = -bf0[6] + bf0[2];
  bf1[7] = -bf0[7] + bf0[3];
Angie Chiang's avatar
Angie Chiang committed
829 830 831 832
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 4
  stage++;
833
  cospi = cospi_arr(cos_bit[stage]);
Angie Chiang's avatar
Angie Chiang committed
834 835 836 837 838 839 840
  bf0 = output;
  bf1 = step;
  bf1[0] = bf0[0];
  bf1[1] = bf0[1];
  bf1[2] = bf0[2];
  bf1[3] = bf0[3];
  bf1[4] = half_btf(cospi[16], bf0[4], cospi[48], bf0[5], cos_bit[stage]);
841
  bf1[5] = half_btf(-cospi[16], bf0[5], cospi[48], bf0[4], cos_bit[stage]);
Angie Chiang's avatar
Angie Chiang committed
842
  bf1[6] = half_btf(-cospi[48], bf0[6], cospi[16], bf0[7], cos_bit[stage]);
843
  bf1[7] = half_btf(cospi[48], bf0[7], cospi[16], bf0[6], cos_bit[stage]);
Angie Chiang's avatar
Angie Chiang committed
844 845 846 847 848 849
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 5
  stage++;
  bf0 = step;
  bf1 = output;
850 851 852 853 854 855 856 857
  bf1[0] = bf0[0] + bf0[2];
  bf1[1] = bf0[1] + bf0[3];
  bf1[2] = -bf0[2] + bf0[0];
  bf1[3] = -bf0[3] + bf0[1];
  bf1[4] = bf0[4] + bf0[6];
  bf1[5] = bf0[5] + bf0[7];
  bf1[6] = -bf0[6] + bf0[4];
  bf1[7] = -bf0[7] + bf0[5];
Angie Chiang's avatar
Angie Chiang committed
858 859 860 861
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 6
  stage++;
862
  cospi = cospi_arr(cos_bit[stage]);
Angie Chiang's avatar
Angie Chiang committed
863 864
  bf0 = output;
  bf1 = step;
865 866 867 868 869 870 871 872
  bf1[0] = bf0[0];
  bf1[1] = bf0[1];
  bf1[2] = half_btf(cospi[32], bf0[2], cospi[32], bf0[3], cos_bit[stage]);
  bf1[3] = half_btf(-cospi[32], bf0[3], cospi[32], bf0[2], cos_bit[stage]);
  bf1[4] = bf0[4];
  bf1[5] = bf0[5];
  bf1[6] = half_btf(cospi[32], bf0[6], cospi[32], bf0[7], cos_bit[stage]);
  bf1[7] = half_btf(-cospi[32], bf0[7], cospi[32], bf0[6], cos_bit[stage]);
Angie Chiang's avatar
Angie Chiang committed
873 874 875 876 877 878
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 7
  stage++;
  bf0 = step;
  bf1 = output;
879 880 881 882 883 884 885 886
  bf1[0] = bf0[0];
  bf1[1] = -bf0[4];
  bf1[2] = bf0[6];
  bf1[3] = -bf0[2];
  bf1[4] = bf0[3];
  bf1[5] = -bf0[7];
  bf1[6] = bf0[5];
  bf1[7] = -bf0[1];
Angie Chiang's avatar
Angie Chiang committed
887 888
  range_check(stage, input, bf1, size, stage_range[stage]);
}
Angie Chiang's avatar
Angie Chiang committed
889

Yaowu Xu's avatar
Yaowu Xu committed
890 891
void av1_iadst16_new(const int32_t *input, int32_t *output,
                     const int8_t *cos_bit, const int8_t *stage_range) {
Angie Chiang's avatar
Angie Chiang committed
892 893 894 895 896 897 898 899 900 901 902 903 904
  const int32_t size = 16;
  const int32_t *cospi;

  int32_t stage = 0;
  int32_t *bf0, *bf1;
  int32_t step[16];

  // stage 0;
  range_check(stage, input, input, size, stage_range[stage]);

  // stage 1;
  stage++;
  bf1 = output;
905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920
  bf1[0] = input[15];
  bf1[1] = input[0];
  bf1[2] = input[13];
  bf1[3] = input[2];
  bf1[4] = input[11];
  bf1[5] = input[4];
  bf1[6] = input[9];
  bf1[7] = input[6];
  bf1[8] = input[7];
  bf1[9] = input[8];
  bf1[10] = input[5];
  bf1[11] = input[10];
  bf1[12] = input[3];
  bf1[13] = input[12];
  bf1[14] = input[1];
  bf1[15] = input[14];
Angie Chiang's avatar
Angie Chiang committed
921 922 923 924
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 2
  stage++;
925
  cospi = cospi_arr(cos_bit[stage]);
Angie Chiang's avatar
Angie Chiang committed
926 927
  bf0 = output;
  bf1 = step;
928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943
  bf1[0] = half_btf(cospi[2], bf0[0], cospi[62], bf0[1], cos_bit[stage]);
  bf1[1] = half_btf(-cospi[2], bf0[1], cospi[62], bf0[0], cos_bit[stage]);
  bf1[2] = half_btf(cospi[10], bf0[2], cospi[54], bf0[3], cos_bit[stage]);
  bf1[3] = half_btf(-cospi[10], bf0[3], cospi[54], bf0[2], cos_bit[stage]);
  bf1[4] = half_btf(cospi[18], bf0[4], cospi[46], bf0[5], cos_bit[stage]);
  bf1[5] = half_btf(-cospi[18], bf0[5], cospi[46], bf0[4], cos_bit[stage]);
  bf1[6] = half_btf(cospi[26], bf0[6], cospi[38], bf0[7], cos_bit[stage]);
  bf1[7] = half_btf(-cospi[26], bf0[7], cospi[38], bf0[6], cos_bit[stage]);
  bf1[8] = half_btf(cospi[34], bf0[8], cospi[30], bf0[9], cos_bit[stage]);
  bf1[9] = half_btf(-cospi[34], bf0[9], cospi[30], bf0[8], cos_bit[stage]);
  bf1[10] = half_btf(cospi[42], bf0[10], cospi[22], bf0[11], cos_bit[stage]);
  bf1[11] = half_btf(-cospi[42], bf0[11], cospi[22], bf0[10], cos_bit[stage]);
  bf1[12] = half_btf(cospi[50], bf0[12], cospi[14], bf0[13], cos_bit[stage]);
  bf1[13] = half_btf(-cospi[50], bf0[13], cospi[14], bf0[12], cos_bit[stage]);
  bf1[14] = half_btf(cospi[58], bf0[14], cospi[6], bf0[15], cos_bit[stage]);
  bf1[15] = half_btf(-cospi[58], bf0[15], cospi[6], bf0[14], cos_bit[stage]);
Angie Chiang's avatar
Angie Chiang committed
944 945 946 947 948 949
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 3
  stage++;
  bf0 = step;
  bf1 = output;
950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965
  bf1[0] = bf0[0] + bf0[8];
  bf1[1] = bf0[1] + bf0[9];
  bf1[2] = bf0[2] + bf0[10];
  bf1[3] = bf0[3] + bf0[11];
  bf1[4] = bf0[4] + bf0[12];
  bf1[5] = bf0[5] + bf0[13];
  bf1[6] = bf0[6] + bf0[14];
  bf1[7] = bf0[7] + bf0[15];
  bf1[8] = -bf0[8] + bf0[0];
  bf1[9] = -bf0[9] + bf0[1];
  bf1[10] = -bf0[10] + bf0[2];
  bf1[11] = -bf0[11] + bf0[3];
  bf1[12] = -bf0[12] + bf0[4];
  bf1[13] = -bf0[13] + bf0[5];
  bf1[14] = -bf0[14] + bf0[6];
  bf1[15] = -bf0[15] + bf0[7];
Angie Chiang's avatar
Angie Chiang committed
966 967 968 969
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 4
  stage++;
970
  cospi = cospi_arr(cos_bit[stage]);
Angie Chiang's avatar
Angie Chiang committed
971 972 973 974 975 976
  bf0 = output;
  bf1 = step;
  bf1[0] = bf0[0];
  bf1[1] = bf0[1];
  bf1[2] = bf0[2];
  bf1[3] = bf0[3];
977 978 979 980 981 982 983 984 985 986 987 988
  bf1[4] = bf0[4];
  bf1[5] = bf0[5];
  bf1[6] = bf0[6];
  bf1[7] = bf0[7];
  bf1[8] = half_btf(cospi[8], bf0[8], cospi[56], bf0[9], cos_bit[stage]);
  bf1[9] = half_btf(-cospi[8], bf0[9], cospi[56], bf0[8], cos_bit[stage]);
  bf1[10] = half_btf(cospi[40], bf0[10], cospi[24], bf0[11], cos_bit[stage]);
  bf1[11] = half_btf(-cospi[40], bf0[11], cospi[24], bf0[10], cos_bit[stage]);
  bf1[12] = half_btf(-cospi[56], bf0[12], cospi[8], bf0[13], cos_bit[stage]);
  bf1[13] = half_btf(cospi[56], bf0[13], cospi[8], bf0[12], cos_bit[stage]);
  bf1[14] = half_btf(-cospi[24], bf0[14], cospi[40], bf0[15], cos_bit[stage]);
  bf1[15] = half_btf(cospi[24], bf0[15], cospi[40], bf0[14], cos_bit[stage]);
Angie Chiang's avatar
Angie Chiang committed
989 990 991 992 993 994 995 996 997 998
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 5
  stage++;
  bf0 = step;
  bf1 = output;
  bf1[0] = bf0[0] + bf0[4];
  bf1[1] = bf0[1] + bf0[5];
  bf1[2] = bf0[2] + bf0[6];
  bf1[3] = bf0[3] + bf0[7];
999 1000 1001 1002
  bf1[4] = -bf0[4] + bf0[0];
  bf1[5] = -bf0[5] + bf0[1];
  bf1[6] = -bf0[6] + bf0[2];
  bf1[7] = -bf0[7] + bf0[3];
Angie Chiang's avatar
Angie Chiang committed
1003 1004 1005 1006
  bf1[8] = bf0[8] + bf0[12];
  bf1[9] = bf0[9] + bf0[13];
  bf1[10] = bf0[10] + bf0[14];
  bf1[11] = bf0[11] + bf0[15];
1007 1008 1009 1010
  bf1[12] = -bf0[12] + bf0[8];
  bf1[13] = -bf0[13] + bf0[9];
  bf1[14] = -bf0[14] + bf0[10];
  bf1[15] = -bf0[15] + bf0[11];
Angie Chiang's avatar
Angie Chiang committed
1011 1012 1013 1014
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 6
  stage++;
1015
  cospi = cospi_arr(cos_bit[stage]);
Angie Chiang's avatar
Angie Chiang committed
1016 1017 1018 1019 1020 1021
  bf0 = output;
  bf1 = step;
  bf1[0] = bf0[0];
  bf1[1] = bf0[1];
  bf1[2] = bf0[2];
  bf1[3] = bf0[3];
1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033
  bf1[4] = half_btf(cospi[16], bf0[4], cospi[48], bf0[5], cos_bit[stage]);
  bf1[5] = half_btf(-cospi[16], bf0[5], cospi[48], bf0[4], cos_bit[stage]);
  bf1[6] = half_btf(-cospi[48], bf0[6], cospi[16], bf0[7], cos_bit[stage]);
  bf1[7] = half_btf(cospi[48], bf0[7], cospi[16], bf0[6], cos_bit[stage]);
  bf1[8] = bf0[8];
  bf1[9] = bf0[9];
  bf1[10] = bf0[10];
  bf1[11] = bf0[11];
  bf1[12] = half_btf(cospi[16], bf0[12], cospi[48], bf0[13], cos_bit[stage]);
  bf1[13] = half_btf(-cospi[16], bf0[13], cospi[48], bf0[12], cos_bit[stage]);
  bf1[14] = half_btf(-cospi[48], bf0[14], cospi[16], bf0[15], cos_bit[stage]);
  bf1[15] = half_btf(cospi[48], bf0[15], cospi[16], bf0[14], cos_bit[stage]);
Angie Chiang's avatar
Angie Chiang committed
1034 1035 1036 1037 1038 1039
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 7
  stage++;
  bf0 = step;
  bf1 = output;
1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055
  bf1[0] = bf0[0] + bf0[2];
  bf1[1] = bf0[1] + bf0[3];
  bf1[2] = -bf0[2] + bf0[0];
  bf1[3] = -bf0[3] + bf0[1];
  bf1[4] = bf0[4] + bf0[6];
  bf1[5] = bf0[5] + bf0[7];
  bf1[6] = -bf0[6] + bf0[4];
  bf1[7] = -bf0[7] + bf0[5];
  bf1[8] = bf0[8] + bf0[10];
  bf1[9] = bf0[9] + bf0[11];
  bf1[10] = -bf0[10] + bf0[8];
  bf1[11] = -bf0[11] + bf0[9];
  bf1[12] = bf0[12] + bf0[14];
  bf1[13] = bf0[13] + bf0[15];
  bf1[14] = -bf0[14] + bf0[12];
  bf1[15] = -bf0[15] + bf0[13];
Angie Chiang's avatar
Angie Chiang committed
1056 1057 1058 1059
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 8
  stage++;
1060
  cospi = cospi_arr(cos_bit[stage]);
Angie Chiang's avatar
Angie Chiang committed
1061 1062
  bf0 = output;
  bf1 = step;
1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078
  bf1[0] = bf0[0];
  bf1[1] = bf0[1];
  bf1[2] = half_btf(cospi[32], bf0[2], cospi[32], bf0[3], cos_bit[stage]);
  bf1[3] = half_btf(-cospi[32], bf0[3], cospi[32], bf0[2], cos_bit[stage]);
  bf1[4] = bf0[4];
  bf1[5] = bf0[5];
  bf1[6] = half_btf(cospi[32], bf0[6], cospi[32], bf0[7], cos_bit[stage]);
  bf1[7] = half_btf(-cospi[32], bf0[7], cospi[32], bf0[6], cos_bit[stage]);
  bf1[8] = bf0[8];
  bf1[9] = bf0[9];
  bf1[10] = half_btf(cospi[32], bf0[10], cospi[32], bf0[11], cos_bit[stage]);
  bf1[11] = half_btf(-cospi[32], bf0[11], cospi[32], bf0[10], cos_bit[stage]);
  bf1[12] = bf0[12];
  bf1[13] = bf0[13];
  bf1[14] = half_btf(cospi[32], bf0[14], cospi[32], bf0[15], cos_bit[stage]);
  bf1[15] = half_btf(-cospi[32], bf0[15], cospi[32], bf0[14], cos_bit[stage]);
Angie Chiang's avatar
Angie Chiang committed
1079 1080 1081 1082 1083 1084
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 9
  stage++;
  bf0 = step;
  bf1 = output;
1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100
  bf1[0] = bf0[0];
  bf1[1] = -bf0[8];
  bf1[2] = bf0[12];
  bf1[3] = -bf0[4];
  bf1[4] = bf0[6];
  bf1[5] = -bf0[14];
  bf1[6] = bf0[10];
  bf1[7] = -bf0[2];
  bf1[8] = bf0[3];
  bf1[9] = -bf0[11];
  bf1[10] = bf0[15];
  bf1[11] = -bf0[7];
  bf1[12] = bf0[5];
  bf1[13] = -bf0[13];
  bf1[14] = bf0[9];
  bf1[15] = -bf0[1];
Angie Chiang's avatar
Angie Chiang committed
1101 1102
  range_check(stage, input, bf1, size, stage_range[stage]);
}
Angie Chiang's avatar
Angie Chiang committed
1103

Yaowu Xu's avatar
Yaowu Xu committed
1104 1105
void av1_iadst32_new(const int32_t *input, int32_t *output,
                     const int8_t *cos_bit, const int8_t *stage_range) {
Angie Chiang's avatar
Angie Chiang committed
1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117
  const int32_t size = 32;
  const int32_t *cospi;

  int32_t stage = 0;
  int32_t *bf0, *bf1;
  int32_t step[32];

  // stage 0;
  range_check(stage, input, input, size, stage_range[stage]);

  // stage 1;
  stage++;
1118
  assert(output != input);
Angie Chiang's avatar
Angie Chiang committed
1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155
  bf1 = output;
  bf1[0] = input[0];
  bf1[1] = -input[31];
  bf1[2] = -input[15];
  bf1[3] = input[16];
  bf1[4] = -input[7];
  bf1[5] = input[24];
  bf1[6] = input[8];
  bf1[7] = -input[23];
  bf1[8] = -input[3];
  bf1[9] = input[28];
  bf1[10] = input[12];
  bf1[11] = -input[19];
  bf1[12] = input[4];
  bf1[13] = -input[27];
  bf1[14] = -input[11];
  bf1[15] = input[20];
  bf1[16] = -input[1];
  bf1[17] = input[30];
  bf1[18] = input[14];
  bf1[19] = -input[17];
  bf1[20] = input[6];
  bf1[21] = -input[25];
  bf1[22] = -input[9];
  bf1[23] = input[22];
  bf1[24] = input[2];
  bf1[25] = -input[29];
  bf1[26] = -input[13];
  bf1[27] = input[18];
  bf1[28] = -input[5];
  bf1[29] = input[26];
  bf1[30] = input[10];
  bf1[31] = -input[21];
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 2
  stage++;
1156
  cospi = cospi_arr(cos_bit[stage]);
Angie Chiang's avatar
Angie Chiang committed
1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232
  bf0 = output;
  bf1 = step;
  bf1[0] = bf0[0];
  bf1[1] = bf0[1];
  bf1[2] = half_btf(cospi[32], bf0[2], cospi[32], bf0[3], cos_bit[stage]);
  bf1[3] = half_btf(cospi[32], bf0[2], -cospi[32], bf0[3], cos_bit[stage]);
  bf1[4] = bf0[4];
  bf1[5] = bf0[5];
  bf1[6] = half_btf(cospi[32], bf0[6], cospi[32], bf0[7], cos_bit[stage]);
  bf1[7] = half_btf(cospi[32], bf0[6], -cospi[32], bf0[7], cos_bit[stage]);
  bf1[8] = bf0[8];
  bf1[9] = bf0[9];
  bf1[10] = half_btf(cospi[32], bf0[10], cospi[32], bf0[11], cos_bit[stage]);
  bf1[11] = half_btf(cospi[32], bf0[10], -cospi[32], bf0[11], cos_bit[stage]);
  bf1[12] = bf0[12];
  bf1[13] = bf0[13];
  bf1[14] = half_btf(cospi[32], bf0[14], cospi[32], bf0[15], cos_bit[stage]);
  bf1[15] = half_btf(cospi[32], bf0[14], -cospi[32], bf0[15], cos_bit[stage]);
  bf1[16] = bf0[16];
  bf1[17] = bf0[17];
  bf1[18] = half_btf(cospi[32], bf0[18], cospi[32], bf0[19], cos_bit[stage]);
  bf1[19] = half_btf(cospi[32], bf0[18], -cospi[32], bf0[19], cos_bit[stage]);
  bf1[20] = bf0[20];
  bf1[21] = bf0[21];
  bf1[22] = half_btf(cospi[32], bf0[22], cospi[32], bf0[23], cos_bit[stage]);
  bf1[23] = half_btf(cospi[32], bf0[22], -cospi[32], bf0[23], cos_bit[stage]);
  bf1[24] = bf0[24];
  bf1[25] = bf0[25];
  bf1[26] = half_btf(cospi[32], bf0[26], cospi[32], bf0[27], cos_bit[stage]);
  bf1[27] = half_btf(cospi[32], bf0[26], -cospi[32], bf0[27], cos_bit[stage]);
  bf1[28] = bf0[28];
  bf1[29] = bf0[29];
  bf1[30] = half_btf(cospi[32], bf0[30], cospi[32], bf0[31], cos_bit[stage]);
  bf1[31] = half_btf(cospi[32], bf0[30], -cospi[32], bf0[31], cos_bit[stage]);
  range_check(stage, input, bf1, size, stage_range[stage]);

  // stage 3
  stage++;
  bf0 = step;
  bf1 = output;
  bf1[0] = bf0[0] + bf0[2];
  bf1[1] = bf0[1] + bf0[3];
  bf1[2] = bf0[0] - bf0[2];
  bf1[3] = bf0[1] - bf0[3];
  bf1[4] = bf0[4] + bf0[6];
  bf1[5] = bf0[5] + bf0[7];
  bf1[6] = bf0[4] - bf0[6];
  bf1[7] = bf0[5] - bf0[7];
  bf1[8] = bf0[8] + bf0[10];
  bf1[9] = bf0[9] + bf0[11];
  bf1[10] = bf0[8] - bf0[10];
  bf1[11] = bf0[9] - bf0[11];
  bf1[<