dct.c 80 KB
Newer Older
Jingning Han's avatar
Jingning Han committed
1
/*
Yaowu Xu's avatar
Yaowu Xu committed
2
 * Copyright (c) 2016, Alliance for Open Media. All rights reserved
Jingning Han's avatar
Jingning Han committed
3
 *
Yaowu Xu's avatar
Yaowu Xu committed
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.
Jingning Han's avatar
Jingning Han committed
10 11 12 13 14
 */

#include <assert.h>
#include <math.h>

Yaowu Xu's avatar
Yaowu Xu committed
15 16
#include "./aom_config.h"
#include "./aom_dsp_rtcd.h"
Geza Lore's avatar
Geza Lore committed
17
#include "./av1_rtcd.h"
18 19
#include "aom_dsp/fwd_txfm.h"
#include "aom_ports/mem.h"
Geza Lore's avatar
Geza Lore committed
20
#include "av1/common/blockd.h"
21
#include "av1/common/av1_fwd_txfm1d.h"
22
#include "av1/common/av1_fwd_txfm1d_cfg.h"
Geza Lore's avatar
Geza Lore committed
23
#include "av1/common/idct.h"
24 25 26
#if CONFIG_DAALA_DCT4
#include "av1/common/daala_tx.h"
#endif
Jingning Han's avatar
Jingning Han committed
27

28 29
static INLINE void range_check(const tran_low_t *input, const int size,
                               const int bit) {
30 31 32 33
#if 0  // CONFIG_COEFFICIENT_RANGE_CHECKING
// TODO(angiebird): the range_check is not used because the bit range
// in fdct# is not correct. Since we are going to merge in a new version
// of fdct# from nextgenv2, we won't fix the incorrect bit range now.
34 35 36 37 38 39 40 41 42 43 44
  int i;
  for (i = 0; i < size; ++i) {
    assert(abs(input[i]) < (1 << bit));
  }
#else
  (void)input;
  (void)size;
  (void)bit;
#endif
}

45 46 47 48 49 50 51 52 53 54 55 56
#if CONFIG_DAALA_DCT4
static void fdct4(const tran_low_t *input, tran_low_t *output) {
  int i;
  od_coeff x[4];
  od_coeff y[4];
  for (i = 0; i < 4; i++) x[i] = (od_coeff)input[i];
  od_bin_fdct4(y, x, 1);
  for (i = 0; i < 4; i++) output[i] = (tran_low_t)y[i];
}

#else

Jingning Han's avatar
Jingning Han committed
57
static void fdct4(const tran_low_t *input, tran_low_t *output) {
58 59 60 61
  tran_high_t temp;
  tran_low_t step[4];

  // stage 0
62
  range_check(input, 4, 14);
63 64 65 66 67 68 69

  // stage 1
  output[0] = input[0] + input[3];
  output[1] = input[1] + input[2];
  output[2] = input[1] - input[2];
  output[3] = input[0] - input[3];

70
  range_check(output, 4, 15);
71 72 73 74 75 76 77 78 79 80 81

  // stage 2
  temp = output[0] * cospi_16_64 + output[1] * cospi_16_64;
  step[0] = (tran_low_t)fdct_round_shift(temp);
  temp = output[1] * -cospi_16_64 + output[0] * cospi_16_64;
  step[1] = (tran_low_t)fdct_round_shift(temp);
  temp = output[2] * cospi_24_64 + output[3] * cospi_8_64;
  step[2] = (tran_low_t)fdct_round_shift(temp);
  temp = output[3] * cospi_24_64 + output[2] * -cospi_8_64;
  step[3] = (tran_low_t)fdct_round_shift(temp);

82
  range_check(step, 4, 16);
83 84 85 86 87 88 89

  // stage 3
  output[0] = step[0];
  output[1] = step[2];
  output[2] = step[1];
  output[3] = step[3];

90
  range_check(output, 4, 16);
Jingning Han's avatar
Jingning Han committed
91
}
92
#endif
Jingning Han's avatar
Jingning Han committed
93 94

static void fdct8(const tran_low_t *input, tran_low_t *output) {
95 96 97 98
  tran_high_t temp;
  tran_low_t step[8];

  // stage 0
99
  range_check(input, 8, 13);
Jingning Han's avatar
Jingning Han committed
100 101

  // stage 1
102 103 104 105 106 107 108 109 110
  output[0] = input[0] + input[7];
  output[1] = input[1] + input[6];
  output[2] = input[2] + input[5];
  output[3] = input[3] + input[4];
  output[4] = input[3] - input[4];
  output[5] = input[2] - input[5];
  output[6] = input[1] - input[6];
  output[7] = input[0] - input[7];

111
  range_check(output, 8, 14);
112 113 114 115 116 117 118 119 120 121 122 123 124

  // stage 2
  step[0] = output[0] + output[3];
  step[1] = output[1] + output[2];
  step[2] = output[1] - output[2];
  step[3] = output[0] - output[3];
  step[4] = output[4];
  temp = output[5] * -cospi_16_64 + output[6] * cospi_16_64;
  step[5] = (tran_low_t)fdct_round_shift(temp);
  temp = output[6] * cospi_16_64 + output[5] * cospi_16_64;
  step[6] = (tran_low_t)fdct_round_shift(temp);
  step[7] = output[7];

125
  range_check(step, 8, 15);
126 127 128 129 130 131 132 133 134 135 136 137 138 139 140

  // stage 3
  temp = step[0] * cospi_16_64 + step[1] * cospi_16_64;
  output[0] = (tran_low_t)fdct_round_shift(temp);
  temp = step[1] * -cospi_16_64 + step[0] * cospi_16_64;
  output[1] = (tran_low_t)fdct_round_shift(temp);
  temp = step[2] * cospi_24_64 + step[3] * cospi_8_64;
  output[2] = (tran_low_t)fdct_round_shift(temp);
  temp = step[3] * cospi_24_64 + step[2] * -cospi_8_64;
  output[3] = (tran_low_t)fdct_round_shift(temp);
  output[4] = step[4] + step[5];
  output[5] = step[4] - step[5];
  output[6] = step[7] - step[6];
  output[7] = step[7] + step[6];

141
  range_check(output, 8, 16);
142 143 144 145 146 147 148 149 150 151 152 153 154 155 156

  // stage 4
  step[0] = output[0];
  step[1] = output[1];
  step[2] = output[2];
  step[3] = output[3];
  temp = output[4] * cospi_28_64 + output[7] * cospi_4_64;
  step[4] = (tran_low_t)fdct_round_shift(temp);
  temp = output[5] * cospi_12_64 + output[6] * cospi_20_64;
  step[5] = (tran_low_t)fdct_round_shift(temp);
  temp = output[6] * cospi_12_64 + output[5] * -cospi_20_64;
  step[6] = (tran_low_t)fdct_round_shift(temp);
  temp = output[7] * cospi_28_64 + output[4] * -cospi_4_64;
  step[7] = (tran_low_t)fdct_round_shift(temp);

157
  range_check(step, 8, 16);
158 159 160 161 162 163 164 165 166 167 168

  // stage 5
  output[0] = step[0];
  output[1] = step[4];
  output[2] = step[2];
  output[3] = step[6];
  output[4] = step[1];
  output[5] = step[5];
  output[6] = step[3];
  output[7] = step[7];

169
  range_check(output, 8, 16);
Jingning Han's avatar
Jingning Han committed
170 171
}

172 173 174
static void fdct16(const tran_low_t *input, tran_low_t *output) {
  tran_high_t temp;
  tran_low_t step[16];
Jingning Han's avatar
Jingning Han committed
175

176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347
  // stage 0
  range_check(input, 16, 13);

  // stage 1
  output[0] = input[0] + input[15];
  output[1] = input[1] + input[14];
  output[2] = input[2] + input[13];
  output[3] = input[3] + input[12];
  output[4] = input[4] + input[11];
  output[5] = input[5] + input[10];
  output[6] = input[6] + input[9];
  output[7] = input[7] + input[8];
  output[8] = input[7] - input[8];
  output[9] = input[6] - input[9];
  output[10] = input[5] - input[10];
  output[11] = input[4] - input[11];
  output[12] = input[3] - input[12];
  output[13] = input[2] - input[13];
  output[14] = input[1] - input[14];
  output[15] = input[0] - input[15];

  range_check(output, 16, 14);

  // stage 2
  step[0] = output[0] + output[7];
  step[1] = output[1] + output[6];
  step[2] = output[2] + output[5];
  step[3] = output[3] + output[4];
  step[4] = output[3] - output[4];
  step[5] = output[2] - output[5];
  step[6] = output[1] - output[6];
  step[7] = output[0] - output[7];
  step[8] = output[8];
  step[9] = output[9];
  temp = output[10] * -cospi_16_64 + output[13] * cospi_16_64;
  step[10] = (tran_low_t)fdct_round_shift(temp);
  temp = output[11] * -cospi_16_64 + output[12] * cospi_16_64;
  step[11] = (tran_low_t)fdct_round_shift(temp);
  temp = output[12] * cospi_16_64 + output[11] * cospi_16_64;
  step[12] = (tran_low_t)fdct_round_shift(temp);
  temp = output[13] * cospi_16_64 + output[10] * cospi_16_64;
  step[13] = (tran_low_t)fdct_round_shift(temp);
  step[14] = output[14];
  step[15] = output[15];

  range_check(step, 16, 15);

  // stage 3
  output[0] = step[0] + step[3];
  output[1] = step[1] + step[2];
  output[2] = step[1] - step[2];
  output[3] = step[0] - step[3];
  output[4] = step[4];
  temp = step[5] * -cospi_16_64 + step[6] * cospi_16_64;
  output[5] = (tran_low_t)fdct_round_shift(temp);
  temp = step[6] * cospi_16_64 + step[5] * cospi_16_64;
  output[6] = (tran_low_t)fdct_round_shift(temp);
  output[7] = step[7];
  output[8] = step[8] + step[11];
  output[9] = step[9] + step[10];
  output[10] = step[9] - step[10];
  output[11] = step[8] - step[11];
  output[12] = step[15] - step[12];
  output[13] = step[14] - step[13];
  output[14] = step[14] + step[13];
  output[15] = step[15] + step[12];

  range_check(output, 16, 16);

  // stage 4
  temp = output[0] * cospi_16_64 + output[1] * cospi_16_64;
  step[0] = (tran_low_t)fdct_round_shift(temp);
  temp = output[1] * -cospi_16_64 + output[0] * cospi_16_64;
  step[1] = (tran_low_t)fdct_round_shift(temp);
  temp = output[2] * cospi_24_64 + output[3] * cospi_8_64;
  step[2] = (tran_low_t)fdct_round_shift(temp);
  temp = output[3] * cospi_24_64 + output[2] * -cospi_8_64;
  step[3] = (tran_low_t)fdct_round_shift(temp);
  step[4] = output[4] + output[5];
  step[5] = output[4] - output[5];
  step[6] = output[7] - output[6];
  step[7] = output[7] + output[6];
  step[8] = output[8];
  temp = output[9] * -cospi_8_64 + output[14] * cospi_24_64;
  step[9] = (tran_low_t)fdct_round_shift(temp);
  temp = output[10] * -cospi_24_64 + output[13] * -cospi_8_64;
  step[10] = (tran_low_t)fdct_round_shift(temp);
  step[11] = output[11];
  step[12] = output[12];
  temp = output[13] * cospi_24_64 + output[10] * -cospi_8_64;
  step[13] = (tran_low_t)fdct_round_shift(temp);
  temp = output[14] * cospi_8_64 + output[9] * cospi_24_64;
  step[14] = (tran_low_t)fdct_round_shift(temp);
  step[15] = output[15];

  range_check(step, 16, 16);

  // stage 5
  output[0] = step[0];
  output[1] = step[1];
  output[2] = step[2];
  output[3] = step[3];
  temp = step[4] * cospi_28_64 + step[7] * cospi_4_64;
  output[4] = (tran_low_t)fdct_round_shift(temp);
  temp = step[5] * cospi_12_64 + step[6] * cospi_20_64;
  output[5] = (tran_low_t)fdct_round_shift(temp);
  temp = step[6] * cospi_12_64 + step[5] * -cospi_20_64;
  output[6] = (tran_low_t)fdct_round_shift(temp);
  temp = step[7] * cospi_28_64 + step[4] * -cospi_4_64;
  output[7] = (tran_low_t)fdct_round_shift(temp);
  output[8] = step[8] + step[9];
  output[9] = step[8] - step[9];
  output[10] = step[11] - step[10];
  output[11] = step[11] + step[10];
  output[12] = step[12] + step[13];
  output[13] = step[12] - step[13];
  output[14] = step[15] - step[14];
  output[15] = step[15] + step[14];

  range_check(output, 16, 16);

  // stage 6
  step[0] = output[0];
  step[1] = output[1];
  step[2] = output[2];
  step[3] = output[3];
  step[4] = output[4];
  step[5] = output[5];
  step[6] = output[6];
  step[7] = output[7];
  temp = output[8] * cospi_30_64 + output[15] * cospi_2_64;
  step[8] = (tran_low_t)fdct_round_shift(temp);
  temp = output[9] * cospi_14_64 + output[14] * cospi_18_64;
  step[9] = (tran_low_t)fdct_round_shift(temp);
  temp = output[10] * cospi_22_64 + output[13] * cospi_10_64;
  step[10] = (tran_low_t)fdct_round_shift(temp);
  temp = output[11] * cospi_6_64 + output[12] * cospi_26_64;
  step[11] = (tran_low_t)fdct_round_shift(temp);
  temp = output[12] * cospi_6_64 + output[11] * -cospi_26_64;
  step[12] = (tran_low_t)fdct_round_shift(temp);
  temp = output[13] * cospi_22_64 + output[10] * -cospi_10_64;
  step[13] = (tran_low_t)fdct_round_shift(temp);
  temp = output[14] * cospi_14_64 + output[9] * -cospi_18_64;
  step[14] = (tran_low_t)fdct_round_shift(temp);
  temp = output[15] * cospi_30_64 + output[8] * -cospi_2_64;
  step[15] = (tran_low_t)fdct_round_shift(temp);

  range_check(step, 16, 16);

  // stage 7
  output[0] = step[0];
  output[1] = step[8];
  output[2] = step[4];
  output[3] = step[12];
  output[4] = step[2];
  output[5] = step[10];
  output[6] = step[6];
  output[7] = step[14];
  output[8] = step[1];
  output[9] = step[9];
  output[10] = step[5];
  output[11] = step[13];
  output[12] = step[3];
  output[13] = step[11];
  output[14] = step[7];
  output[15] = step[15];

  range_check(output, 16, 16);
}

static void fdct32(const tran_low_t *input, tran_low_t *output) {
  tran_high_t temp;
348
  tran_low_t step[32];
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

  // stage 0
  range_check(input, 32, 14);

  // stage 1
  output[0] = input[0] + input[31];
  output[1] = input[1] + input[30];
  output[2] = input[2] + input[29];
  output[3] = input[3] + input[28];
  output[4] = input[4] + input[27];
  output[5] = input[5] + input[26];
  output[6] = input[6] + input[25];
  output[7] = input[7] + input[24];
  output[8] = input[8] + input[23];
  output[9] = input[9] + input[22];
  output[10] = input[10] + input[21];
  output[11] = input[11] + input[20];
  output[12] = input[12] + input[19];
  output[13] = input[13] + input[18];
  output[14] = input[14] + input[17];
  output[15] = input[15] + input[16];
  output[16] = input[15] - input[16];
  output[17] = input[14] - input[17];
  output[18] = input[13] - input[18];
  output[19] = input[12] - input[19];
  output[20] = input[11] - input[20];
  output[21] = input[10] - input[21];
  output[22] = input[9] - input[22];
  output[23] = input[8] - input[23];
  output[24] = input[7] - input[24];
  output[25] = input[6] - input[25];
  output[26] = input[5] - input[26];
  output[27] = input[4] - input[27];
  output[28] = input[3] - input[28];
  output[29] = input[2] - input[29];
  output[30] = input[1] - input[30];
  output[31] = input[0] - input[31];

  range_check(output, 32, 15);
Jingning Han's avatar
Jingning Han committed
388

389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430
  // stage 2
  step[0] = output[0] + output[15];
  step[1] = output[1] + output[14];
  step[2] = output[2] + output[13];
  step[3] = output[3] + output[12];
  step[4] = output[4] + output[11];
  step[5] = output[5] + output[10];
  step[6] = output[6] + output[9];
  step[7] = output[7] + output[8];
  step[8] = output[7] - output[8];
  step[9] = output[6] - output[9];
  step[10] = output[5] - output[10];
  step[11] = output[4] - output[11];
  step[12] = output[3] - output[12];
  step[13] = output[2] - output[13];
  step[14] = output[1] - output[14];
  step[15] = output[0] - output[15];
  step[16] = output[16];
  step[17] = output[17];
  step[18] = output[18];
  step[19] = output[19];
  temp = output[20] * -cospi_16_64 + output[27] * cospi_16_64;
  step[20] = (tran_low_t)fdct_round_shift(temp);
  temp = output[21] * -cospi_16_64 + output[26] * cospi_16_64;
  step[21] = (tran_low_t)fdct_round_shift(temp);
  temp = output[22] * -cospi_16_64 + output[25] * cospi_16_64;
  step[22] = (tran_low_t)fdct_round_shift(temp);
  temp = output[23] * -cospi_16_64 + output[24] * cospi_16_64;
  step[23] = (tran_low_t)fdct_round_shift(temp);
  temp = output[24] * cospi_16_64 + output[23] * cospi_16_64;
  step[24] = (tran_low_t)fdct_round_shift(temp);
  temp = output[25] * cospi_16_64 + output[22] * cospi_16_64;
  step[25] = (tran_low_t)fdct_round_shift(temp);
  temp = output[26] * cospi_16_64 + output[21] * cospi_16_64;
  step[26] = (tran_low_t)fdct_round_shift(temp);
  temp = output[27] * cospi_16_64 + output[20] * cospi_16_64;
  step[27] = (tran_low_t)fdct_round_shift(temp);
  step[28] = output[28];
  step[29] = output[29];
  step[30] = output[30];
  step[31] = output[31];

431
  range_check(step, 32, 16);
432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516

  // stage 3
  output[0] = step[0] + step[7];
  output[1] = step[1] + step[6];
  output[2] = step[2] + step[5];
  output[3] = step[3] + step[4];
  output[4] = step[3] - step[4];
  output[5] = step[2] - step[5];
  output[6] = step[1] - step[6];
  output[7] = step[0] - step[7];
  output[8] = step[8];
  output[9] = step[9];
  temp = step[10] * -cospi_16_64 + step[13] * cospi_16_64;
  output[10] = (tran_low_t)fdct_round_shift(temp);
  temp = step[11] * -cospi_16_64 + step[12] * cospi_16_64;
  output[11] = (tran_low_t)fdct_round_shift(temp);
  temp = step[12] * cospi_16_64 + step[11] * cospi_16_64;
  output[12] = (tran_low_t)fdct_round_shift(temp);
  temp = step[13] * cospi_16_64 + step[10] * cospi_16_64;
  output[13] = (tran_low_t)fdct_round_shift(temp);
  output[14] = step[14];
  output[15] = step[15];
  output[16] = step[16] + step[23];
  output[17] = step[17] + step[22];
  output[18] = step[18] + step[21];
  output[19] = step[19] + step[20];
  output[20] = step[19] - step[20];
  output[21] = step[18] - step[21];
  output[22] = step[17] - step[22];
  output[23] = step[16] - step[23];
  output[24] = step[31] - step[24];
  output[25] = step[30] - step[25];
  output[26] = step[29] - step[26];
  output[27] = step[28] - step[27];
  output[28] = step[28] + step[27];
  output[29] = step[29] + step[26];
  output[30] = step[30] + step[25];
  output[31] = step[31] + step[24];

  range_check(output, 32, 17);

  // stage 4
  step[0] = output[0] + output[3];
  step[1] = output[1] + output[2];
  step[2] = output[1] - output[2];
  step[3] = output[0] - output[3];
  step[4] = output[4];
  temp = output[5] * -cospi_16_64 + output[6] * cospi_16_64;
  step[5] = (tran_low_t)fdct_round_shift(temp);
  temp = output[6] * cospi_16_64 + output[5] * cospi_16_64;
  step[6] = (tran_low_t)fdct_round_shift(temp);
  step[7] = output[7];
  step[8] = output[8] + output[11];
  step[9] = output[9] + output[10];
  step[10] = output[9] - output[10];
  step[11] = output[8] - output[11];
  step[12] = output[15] - output[12];
  step[13] = output[14] - output[13];
  step[14] = output[14] + output[13];
  step[15] = output[15] + output[12];
  step[16] = output[16];
  step[17] = output[17];
  temp = output[18] * -cospi_8_64 + output[29] * cospi_24_64;
  step[18] = (tran_low_t)fdct_round_shift(temp);
  temp = output[19] * -cospi_8_64 + output[28] * cospi_24_64;
  step[19] = (tran_low_t)fdct_round_shift(temp);
  temp = output[20] * -cospi_24_64 + output[27] * -cospi_8_64;
  step[20] = (tran_low_t)fdct_round_shift(temp);
  temp = output[21] * -cospi_24_64 + output[26] * -cospi_8_64;
  step[21] = (tran_low_t)fdct_round_shift(temp);
  step[22] = output[22];
  step[23] = output[23];
  step[24] = output[24];
  step[25] = output[25];
  temp = output[26] * cospi_24_64 + output[21] * -cospi_8_64;
  step[26] = (tran_low_t)fdct_round_shift(temp);
  temp = output[27] * cospi_24_64 + output[20] * -cospi_8_64;
  step[27] = (tran_low_t)fdct_round_shift(temp);
  temp = output[28] * cospi_8_64 + output[19] * cospi_24_64;
  step[28] = (tran_low_t)fdct_round_shift(temp);
  temp = output[29] * cospi_8_64 + output[18] * cospi_24_64;
  step[29] = (tran_low_t)fdct_round_shift(temp);
  step[30] = output[30];
  step[31] = output[31];

517
  range_check(step, 32, 18);
518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608

  // stage 5
  temp = step[0] * cospi_16_64 + step[1] * cospi_16_64;
  output[0] = (tran_low_t)fdct_round_shift(temp);
  temp = step[1] * -cospi_16_64 + step[0] * cospi_16_64;
  output[1] = (tran_low_t)fdct_round_shift(temp);
  temp = step[2] * cospi_24_64 + step[3] * cospi_8_64;
  output[2] = (tran_low_t)fdct_round_shift(temp);
  temp = step[3] * cospi_24_64 + step[2] * -cospi_8_64;
  output[3] = (tran_low_t)fdct_round_shift(temp);
  output[4] = step[4] + step[5];
  output[5] = step[4] - step[5];
  output[6] = step[7] - step[6];
  output[7] = step[7] + step[6];
  output[8] = step[8];
  temp = step[9] * -cospi_8_64 + step[14] * cospi_24_64;
  output[9] = (tran_low_t)fdct_round_shift(temp);
  temp = step[10] * -cospi_24_64 + step[13] * -cospi_8_64;
  output[10] = (tran_low_t)fdct_round_shift(temp);
  output[11] = step[11];
  output[12] = step[12];
  temp = step[13] * cospi_24_64 + step[10] * -cospi_8_64;
  output[13] = (tran_low_t)fdct_round_shift(temp);
  temp = step[14] * cospi_8_64 + step[9] * cospi_24_64;
  output[14] = (tran_low_t)fdct_round_shift(temp);
  output[15] = step[15];
  output[16] = step[16] + step[19];
  output[17] = step[17] + step[18];
  output[18] = step[17] - step[18];
  output[19] = step[16] - step[19];
  output[20] = step[23] - step[20];
  output[21] = step[22] - step[21];
  output[22] = step[22] + step[21];
  output[23] = step[23] + step[20];
  output[24] = step[24] + step[27];
  output[25] = step[25] + step[26];
  output[26] = step[25] - step[26];
  output[27] = step[24] - step[27];
  output[28] = step[31] - step[28];
  output[29] = step[30] - step[29];
  output[30] = step[30] + step[29];
  output[31] = step[31] + step[28];

  range_check(output, 32, 18);

  // stage 6
  step[0] = output[0];
  step[1] = output[1];
  step[2] = output[2];
  step[3] = output[3];
  temp = output[4] * cospi_28_64 + output[7] * cospi_4_64;
  step[4] = (tran_low_t)fdct_round_shift(temp);
  temp = output[5] * cospi_12_64 + output[6] * cospi_20_64;
  step[5] = (tran_low_t)fdct_round_shift(temp);
  temp = output[6] * cospi_12_64 + output[5] * -cospi_20_64;
  step[6] = (tran_low_t)fdct_round_shift(temp);
  temp = output[7] * cospi_28_64 + output[4] * -cospi_4_64;
  step[7] = (tran_low_t)fdct_round_shift(temp);
  step[8] = output[8] + output[9];
  step[9] = output[8] - output[9];
  step[10] = output[11] - output[10];
  step[11] = output[11] + output[10];
  step[12] = output[12] + output[13];
  step[13] = output[12] - output[13];
  step[14] = output[15] - output[14];
  step[15] = output[15] + output[14];
  step[16] = output[16];
  temp = output[17] * -cospi_4_64 + output[30] * cospi_28_64;
  step[17] = (tran_low_t)fdct_round_shift(temp);
  temp = output[18] * -cospi_28_64 + output[29] * -cospi_4_64;
  step[18] = (tran_low_t)fdct_round_shift(temp);
  step[19] = output[19];
  step[20] = output[20];
  temp = output[21] * -cospi_20_64 + output[26] * cospi_12_64;
  step[21] = (tran_low_t)fdct_round_shift(temp);
  temp = output[22] * -cospi_12_64 + output[25] * -cospi_20_64;
  step[22] = (tran_low_t)fdct_round_shift(temp);
  step[23] = output[23];
  step[24] = output[24];
  temp = output[25] * cospi_12_64 + output[22] * -cospi_20_64;
  step[25] = (tran_low_t)fdct_round_shift(temp);
  temp = output[26] * cospi_20_64 + output[21] * cospi_12_64;
  step[26] = (tran_low_t)fdct_round_shift(temp);
  step[27] = output[27];
  step[28] = output[28];
  temp = output[29] * cospi_28_64 + output[18] * -cospi_4_64;
  step[29] = (tran_low_t)fdct_round_shift(temp);
  temp = output[30] * cospi_4_64 + output[17] * cospi_28_64;
  step[30] = (tran_low_t)fdct_round_shift(temp);
  step[31] = output[31];

609
  range_check(step, 32, 18);
610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704

  // stage 7
  output[0] = step[0];
  output[1] = step[1];
  output[2] = step[2];
  output[3] = step[3];
  output[4] = step[4];
  output[5] = step[5];
  output[6] = step[6];
  output[7] = step[7];
  temp = step[8] * cospi_30_64 + step[15] * cospi_2_64;
  output[8] = (tran_low_t)fdct_round_shift(temp);
  temp = step[9] * cospi_14_64 + step[14] * cospi_18_64;
  output[9] = (tran_low_t)fdct_round_shift(temp);
  temp = step[10] * cospi_22_64 + step[13] * cospi_10_64;
  output[10] = (tran_low_t)fdct_round_shift(temp);
  temp = step[11] * cospi_6_64 + step[12] * cospi_26_64;
  output[11] = (tran_low_t)fdct_round_shift(temp);
  temp = step[12] * cospi_6_64 + step[11] * -cospi_26_64;
  output[12] = (tran_low_t)fdct_round_shift(temp);
  temp = step[13] * cospi_22_64 + step[10] * -cospi_10_64;
  output[13] = (tran_low_t)fdct_round_shift(temp);
  temp = step[14] * cospi_14_64 + step[9] * -cospi_18_64;
  output[14] = (tran_low_t)fdct_round_shift(temp);
  temp = step[15] * cospi_30_64 + step[8] * -cospi_2_64;
  output[15] = (tran_low_t)fdct_round_shift(temp);
  output[16] = step[16] + step[17];
  output[17] = step[16] - step[17];
  output[18] = step[19] - step[18];
  output[19] = step[19] + step[18];
  output[20] = step[20] + step[21];
  output[21] = step[20] - step[21];
  output[22] = step[23] - step[22];
  output[23] = step[23] + step[22];
  output[24] = step[24] + step[25];
  output[25] = step[24] - step[25];
  output[26] = step[27] - step[26];
  output[27] = step[27] + step[26];
  output[28] = step[28] + step[29];
  output[29] = step[28] - step[29];
  output[30] = step[31] - step[30];
  output[31] = step[31] + step[30];

  range_check(output, 32, 18);

  // stage 8
  step[0] = output[0];
  step[1] = output[1];
  step[2] = output[2];
  step[3] = output[3];
  step[4] = output[4];
  step[5] = output[5];
  step[6] = output[6];
  step[7] = output[7];
  step[8] = output[8];
  step[9] = output[9];
  step[10] = output[10];
  step[11] = output[11];
  step[12] = output[12];
  step[13] = output[13];
  step[14] = output[14];
  step[15] = output[15];
  temp = output[16] * cospi_31_64 + output[31] * cospi_1_64;
  step[16] = (tran_low_t)fdct_round_shift(temp);
  temp = output[17] * cospi_15_64 + output[30] * cospi_17_64;
  step[17] = (tran_low_t)fdct_round_shift(temp);
  temp = output[18] * cospi_23_64 + output[29] * cospi_9_64;
  step[18] = (tran_low_t)fdct_round_shift(temp);
  temp = output[19] * cospi_7_64 + output[28] * cospi_25_64;
  step[19] = (tran_low_t)fdct_round_shift(temp);
  temp = output[20] * cospi_27_64 + output[27] * cospi_5_64;
  step[20] = (tran_low_t)fdct_round_shift(temp);
  temp = output[21] * cospi_11_64 + output[26] * cospi_21_64;
  step[21] = (tran_low_t)fdct_round_shift(temp);
  temp = output[22] * cospi_19_64 + output[25] * cospi_13_64;
  step[22] = (tran_low_t)fdct_round_shift(temp);
  temp = output[23] * cospi_3_64 + output[24] * cospi_29_64;
  step[23] = (tran_low_t)fdct_round_shift(temp);
  temp = output[24] * cospi_3_64 + output[23] * -cospi_29_64;
  step[24] = (tran_low_t)fdct_round_shift(temp);
  temp = output[25] * cospi_19_64 + output[22] * -cospi_13_64;
  step[25] = (tran_low_t)fdct_round_shift(temp);
  temp = output[26] * cospi_11_64 + output[21] * -cospi_21_64;
  step[26] = (tran_low_t)fdct_round_shift(temp);
  temp = output[27] * cospi_27_64 + output[20] * -cospi_5_64;
  step[27] = (tran_low_t)fdct_round_shift(temp);
  temp = output[28] * cospi_7_64 + output[19] * -cospi_25_64;
  step[28] = (tran_low_t)fdct_round_shift(temp);
  temp = output[29] * cospi_23_64 + output[18] * -cospi_9_64;
  step[29] = (tran_low_t)fdct_round_shift(temp);
  temp = output[30] * cospi_15_64 + output[17] * -cospi_17_64;
  step[30] = (tran_low_t)fdct_round_shift(temp);
  temp = output[31] * cospi_31_64 + output[16] * -cospi_1_64;
  step[31] = (tran_low_t)fdct_round_shift(temp);

705
  range_check(step, 32, 18);
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

  // stage 9
  output[0] = step[0];
  output[1] = step[16];
  output[2] = step[8];
  output[3] = step[24];
  output[4] = step[4];
  output[5] = step[20];
  output[6] = step[12];
  output[7] = step[28];
  output[8] = step[2];
  output[9] = step[18];
  output[10] = step[10];
  output[11] = step[26];
  output[12] = step[6];
  output[13] = step[22];
  output[14] = step[14];
  output[15] = step[30];
  output[16] = step[1];
  output[17] = step[17];
  output[18] = step[9];
  output[19] = step[25];
  output[20] = step[5];
  output[21] = step[21];
  output[22] = step[13];
  output[23] = step[29];
  output[24] = step[3];
  output[25] = step[19];
  output[26] = step[11];
  output[27] = step[27];
  output[28] = step[7];
  output[29] = step[23];
  output[30] = step[15];
  output[31] = step[31];

  range_check(output, 32, 18);
Jingning Han's avatar
Jingning Han committed
742 743
}

744 745
#ifndef AV1_DCT_GTEST

Jingning Han's avatar
Jingning Han committed
746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798
static void fadst4(const tran_low_t *input, tran_low_t *output) {
  tran_high_t x0, x1, x2, x3;
  tran_high_t s0, s1, s2, s3, s4, s5, s6, s7;

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

  if (!(x0 | x1 | x2 | x3)) {
    output[0] = output[1] = output[2] = output[3] = 0;
    return;
  }

  s0 = sinpi_1_9 * x0;
  s1 = sinpi_4_9 * x0;
  s2 = sinpi_2_9 * x1;
  s3 = sinpi_1_9 * x1;
  s4 = sinpi_3_9 * x2;
  s5 = sinpi_4_9 * x3;
  s6 = sinpi_2_9 * x3;
  s7 = x0 + x1 - x3;

  x0 = s0 + s2 + s5;
  x1 = sinpi_3_9 * s7;
  x2 = s1 - s3 + s6;
  x3 = s4;

  s0 = x0 + x3;
  s1 = x1;
  s2 = x2 - x3;
  s3 = x2 - x0 + x3;

  // 1-D transform scaling factor is sqrt(2).
  output[0] = (tran_low_t)fdct_round_shift(s0);
  output[1] = (tran_low_t)fdct_round_shift(s1);
  output[2] = (tran_low_t)fdct_round_shift(s2);
  output[3] = (tran_low_t)fdct_round_shift(s3);
}

static void fadst8(const tran_low_t *input, tran_low_t *output) {
  tran_high_t s0, s1, s2, s3, s4, s5, s6, s7;

  tran_high_t x0 = input[7];
  tran_high_t x1 = input[0];
  tran_high_t x2 = input[5];
  tran_high_t x3 = input[2];
  tran_high_t x4 = input[3];
  tran_high_t x5 = input[4];
  tran_high_t x6 = input[1];
  tran_high_t x7 = input[6];

  // stage 1
799 800
  s0 = cospi_2_64 * x0 + cospi_30_64 * x1;
  s1 = cospi_30_64 * x0 - cospi_2_64 * x1;
Jingning Han's avatar
Jingning Han committed
801 802 803 804
  s2 = cospi_10_64 * x2 + cospi_22_64 * x3;
  s3 = cospi_22_64 * x2 - cospi_10_64 * x3;
  s4 = cospi_18_64 * x4 + cospi_14_64 * x5;
  s5 = cospi_14_64 * x4 - cospi_18_64 * x5;
805 806
  s6 = cospi_26_64 * x6 + cospi_6_64 * x7;
  s7 = cospi_6_64 * x6 - cospi_26_64 * x7;
Jingning Han's avatar
Jingning Han committed
807

808 809 810 811
  x0 = s0 + s4;
  x1 = s1 + s5;
  x2 = s2 + s6;
  x3 = s3 + s7;
Jingning Han's avatar
Jingning Han committed
812 813 814 815 816 817 818 819 820 821
  x4 = fdct_round_shift(s0 - s4);
  x5 = fdct_round_shift(s1 - s5);
  x6 = fdct_round_shift(s2 - s6);
  x7 = fdct_round_shift(s3 - s7);

  // stage 2
  s0 = x0;
  s1 = x1;
  s2 = x2;
  s3 = x3;
822 823 824 825
  s4 = cospi_8_64 * x4 + cospi_24_64 * x5;
  s5 = cospi_24_64 * x4 - cospi_8_64 * x5;
  s6 = -cospi_24_64 * x6 + cospi_8_64 * x7;
  s7 = cospi_8_64 * x6 + cospi_24_64 * x7;
Jingning Han's avatar
Jingning Han committed
826

827 828 829 830
  x0 = fdct_round_shift(s0 + s2);
  x1 = fdct_round_shift(s1 + s3);
  x2 = fdct_round_shift(s0 - s2);
  x3 = fdct_round_shift(s1 - s3);
Jingning Han's avatar
Jingning Han committed
831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878
  x4 = fdct_round_shift(s4 + s6);
  x5 = fdct_round_shift(s5 + s7);
  x6 = fdct_round_shift(s4 - s6);
  x7 = fdct_round_shift(s5 - s7);

  // stage 3
  s2 = cospi_16_64 * (x2 + x3);
  s3 = cospi_16_64 * (x2 - x3);
  s6 = cospi_16_64 * (x6 + x7);
  s7 = cospi_16_64 * (x6 - x7);

  x2 = fdct_round_shift(s2);
  x3 = fdct_round_shift(s3);
  x6 = fdct_round_shift(s6);
  x7 = fdct_round_shift(s7);

  output[0] = (tran_low_t)x0;
  output[1] = (tran_low_t)-x4;
  output[2] = (tran_low_t)x6;
  output[3] = (tran_low_t)-x2;
  output[4] = (tran_low_t)x3;
  output[5] = (tran_low_t)-x7;
  output[6] = (tran_low_t)x5;
  output[7] = (tran_low_t)-x1;
}

static void fadst16(const tran_low_t *input, tran_low_t *output) {
  tran_high_t s0, s1, s2, s3, s4, s5, s6, s7, s8;
  tran_high_t s9, s10, s11, s12, s13, s14, s15;

  tran_high_t x0 = input[15];
  tran_high_t x1 = input[0];
  tran_high_t x2 = input[13];
  tran_high_t x3 = input[2];
  tran_high_t x4 = input[11];
  tran_high_t x5 = input[4];
  tran_high_t x6 = input[9];
  tran_high_t x7 = input[6];
  tran_high_t x8 = input[7];
  tran_high_t x9 = input[8];
  tran_high_t x10 = input[5];
  tran_high_t x11 = input[10];
  tran_high_t x12 = input[3];
  tran_high_t x13 = input[12];
  tran_high_t x14 = input[1];
  tran_high_t x15 = input[14];

  // stage 1
879
  s0 = x0 * cospi_1_64 + x1 * cospi_31_64;
Jingning Han's avatar
Jingning Han committed
880
  s1 = x0 * cospi_31_64 - x1 * cospi_1_64;
881
  s2 = x2 * cospi_5_64 + x3 * cospi_27_64;
Jingning Han's avatar
Jingning Han committed
882
  s3 = x2 * cospi_27_64 - x3 * cospi_5_64;
883
  s4 = x4 * cospi_9_64 + x5 * cospi_23_64;
Jingning Han's avatar
Jingning Han committed
884 885 886 887 888 889 890 891
  s5 = x4 * cospi_23_64 - x5 * cospi_9_64;
  s6 = x6 * cospi_13_64 + x7 * cospi_19_64;
  s7 = x6 * cospi_19_64 - x7 * cospi_13_64;
  s8 = x8 * cospi_17_64 + x9 * cospi_15_64;
  s9 = x8 * cospi_15_64 - x9 * cospi_17_64;
  s10 = x10 * cospi_21_64 + x11 * cospi_11_64;
  s11 = x10 * cospi_11_64 - x11 * cospi_21_64;
  s12 = x12 * cospi_25_64 + x13 * cospi_7_64;
892
  s13 = x12 * cospi_7_64 - x13 * cospi_25_64;
Jingning Han's avatar
Jingning Han committed
893
  s14 = x14 * cospi_29_64 + x15 * cospi_3_64;
894
  s15 = x14 * cospi_3_64 - x15 * cospi_29_64;
Jingning Han's avatar
Jingning Han committed
895

896 897 898 899 900 901 902 903 904
  x0 = s0 + s8;
  x1 = s1 + s9;
  x2 = s2 + s10;
  x3 = s3 + s11;
  x4 = s4 + s12;
  x5 = s5 + s13;
  x6 = s6 + s14;
  x7 = s7 + s15;

905 906
  x8 = fdct_round_shift(s0 - s8);
  x9 = fdct_round_shift(s1 - s9);
Jingning Han's avatar
Jingning Han committed
907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922
  x10 = fdct_round_shift(s2 - s10);
  x11 = fdct_round_shift(s3 - s11);
  x12 = fdct_round_shift(s4 - s12);
  x13 = fdct_round_shift(s5 - s13);
  x14 = fdct_round_shift(s6 - s14);
  x15 = fdct_round_shift(s7 - s15);

  // stage 2
  s0 = x0;
  s1 = x1;
  s2 = x2;
  s3 = x3;
  s4 = x4;
  s5 = x5;
  s6 = x6;
  s7 = x7;
923 924 925 926 927 928 929 930
  s8 = x8 * cospi_4_64 + x9 * cospi_28_64;
  s9 = x8 * cospi_28_64 - x9 * cospi_4_64;
  s10 = x10 * cospi_20_64 + x11 * cospi_12_64;
  s11 = x10 * cospi_12_64 - x11 * cospi_20_64;
  s12 = -x12 * cospi_28_64 + x13 * cospi_4_64;
  s13 = x12 * cospi_4_64 + x13 * cospi_28_64;
  s14 = -x14 * cospi_12_64 + x15 * cospi_20_64;
  s15 = x14 * cospi_20_64 + x15 * cospi_12_64;
Jingning Han's avatar
Jingning Han committed
931 932 933 934 935

  x0 = s0 + s4;
  x1 = s1 + s5;
  x2 = s2 + s6;
  x3 = s3 + s7;
936 937 938 939 940 941 942 943 944
  x4 = fdct_round_shift(s0 - s4);
  x5 = fdct_round_shift(s1 - s5);
  x6 = fdct_round_shift(s2 - s6);
  x7 = fdct_round_shift(s3 - s7);

  x8 = s8 + s12;
  x9 = s9 + s13;
  x10 = s10 + s14;
  x11 = s11 + s15;
Jingning Han's avatar
Jingning Han committed
945 946 947 948 949 950 951 952 953 954
  x12 = fdct_round_shift(s8 - s12);
  x13 = fdct_round_shift(s9 - s13);
  x14 = fdct_round_shift(s10 - s14);
  x15 = fdct_round_shift(s11 - s15);

  // stage 3
  s0 = x0;
  s1 = x1;
  s2 = x2;
  s3 = x3;
955
  s4 = x4 * cospi_8_64 + x5 * cospi_24_64;
Jingning Han's avatar
Jingning Han committed
956
  s5 = x4 * cospi_24_64 - x5 * cospi_8_64;
957 958
  s6 = -x6 * cospi_24_64 + x7 * cospi_8_64;
  s7 = x6 * cospi_8_64 + x7 * cospi_24_64;
Jingning Han's avatar
Jingning Han committed
959 960 961 962
  s8 = x8;
  s9 = x9;
  s10 = x10;
  s11 = x11;
963
  s12 = x12 * cospi_8_64 + x13 * cospi_24_64;
Jingning Han's avatar
Jingning Han committed
964
  s13 = x12 * cospi_24_64 - x13 * cospi_8_64;
965 966
  s14 = -x14 * cospi_24_64 + x15 * cospi_8_64;
  s15 = x14 * cospi_8_64 + x15 * cospi_24_64;
Jingning Han's avatar
Jingning Han committed
967

968 969 970 971 972
  x0 = fdct_round_shift(s0 + s2);
  x1 = fdct_round_shift(s1 + s3);
  x2 = fdct_round_shift(s0 - s2);
  x3 = fdct_round_shift(s1 - s3);

Jingning Han's avatar
Jingning Han committed
973 974 975 976
  x4 = fdct_round_shift(s4 + s6);
  x5 = fdct_round_shift(s5 + s7);
  x6 = fdct_round_shift(s4 - s6);
  x7 = fdct_round_shift(s5 - s7);
977 978 979 980 981 982

  x8 = fdct_round_shift(s8 + s10);
  x9 = fdct_round_shift(s9 + s11);
  x10 = fdct_round_shift(s8 - s10);
  x11 = fdct_round_shift(s9 - s11);

Jingning Han's avatar
Jingning Han committed
983 984 985 986 987 988
  x12 = fdct_round_shift(s12 + s14);
  x13 = fdct_round_shift(s13 + s15);
  x14 = fdct_round_shift(s12 - s14);
  x15 = fdct_round_shift(s13 - s15);

  // stage 4
989
  s2 = (-cospi_16_64) * (x2 + x3);
Jingning Han's avatar
Jingning Han committed
990 991
  s3 = cospi_16_64 * (x2 - x3);
  s6 = cospi_16_64 * (x6 + x7);
992
  s7 = cospi_16_64 * (-x6 + x7);
Jingning Han's avatar
Jingning Han committed
993
  s10 = cospi_16_64 * (x10 + x11);
994 995
  s11 = cospi_16_64 * (-x10 + x11);
  s14 = (-cospi_16_64) * (x14 + x15);
Jingning Han's avatar
Jingning Han committed
996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024
  s15 = cospi_16_64 * (x14 - x15);

  x2 = fdct_round_shift(s2);
  x3 = fdct_round_shift(s3);
  x6 = fdct_round_shift(s6);
  x7 = fdct_round_shift(s7);
  x10 = fdct_round_shift(s10);
  x11 = fdct_round_shift(s11);
  x14 = fdct_round_shift(s14);
  x15 = fdct_round_shift(s15);

  output[0] = (tran_low_t)x0;
  output[1] = (tran_low_t)-x8;
  output[2] = (tran_low_t)x12;
  output[3] = (tran_low_t)-x4;
  output[4] = (tran_low_t)x6;
  output[5] = (tran_low_t)x14;
  output[6] = (tran_low_t)x10;
  output[7] = (tran_low_t)x2;
  output[8] = (tran_low_t)x3;
  output[9] = (tran_low_t)x11;
  output[10] = (tran_low_t)x15;
  output[11] = (tran_low_t)x7;
  output[12] = (tran_low_t)x5;
  output[13] = (tran_low_t)-x13;
  output[14] = (tran_low_t)x9;
  output[15] = (tran_low_t)-x1;
}

1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039
// For use in lieu of ADST
static void fhalfright32(const tran_low_t *input, tran_low_t *output) {
  int i;
  tran_low_t inputhalf[16];
  for (i = 0; i < 16; ++i) {
    output[16 + i] = input[i] * 4;
  }
  // Multiply input by sqrt(2)
  for (i = 0; i < 16; ++i) {
    inputhalf[i] = (tran_low_t)fdct_round_shift(input[i + 16] * Sqrt2);
  }
  fdct16(inputhalf, output);
  // Note overall scaling factor is 4 times orthogonal
}

Sarah Parker's avatar
Sarah Parker committed
1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054
#if CONFIG_MRC_TX
static void get_masked_residual32_fwd(const tran_low_t *input,
                                      tran_low_t *output) {
  // placeholder for future bitmask creation
  memcpy(output, input, 32 * 32 * sizeof(*input));
}

static void fmrc32(const tran_low_t *input, tran_low_t *output) {
  // placeholder for mrc_dct, this just performs regular dct
  tran_low_t masked_input[32 * 32];
  get_masked_residual32_fwd(input, masked_input);
  fdct32(masked_input, output);
}
#endif  // CONFIG_MRC_TX

Lester Lu's avatar
Lester Lu committed
1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081
#if CONFIG_LGT
static void flgt4(const tran_low_t *input, tran_low_t *output,
                  const tran_high_t *lgtmtx) {
  if (!(input[0] | input[1] | input[2] | input[3])) {
    output[0] = output[1] = output[2] = output[3] = 0;
    return;
  }

  // evaluate s[j] = sum of all lgtmtx[j][i]*input[i] over i=1,...,4
  tran_high_t s[4] = { 0 };
  for (int i = 0; i < 4; ++i)
    for (int j = 0; j < 4; ++j) s[j] += lgtmtx[j * 4 + i] * input[i];

  for (int i = 0; i < 4; ++i) output[i] = (tran_low_t)fdct_round_shift(s[i]);
}

static void flgt8(const tran_low_t *input, tran_low_t *output,
                  const tran_high_t *lgtmtx) {
  // evaluate s[j] = sum of all lgtmtx[j][i]*input[i] over i=1,...,8
  tran_high_t s[8] = { 0 };
  for (int i = 0; i < 8; ++i)
    for (int j = 0; j < 8; ++j) s[j] += lgtmtx[j * 8 + i] * input[i];

  for (int i = 0; i < 8; ++i) output[i] = (tran_low_t)fdct_round_shift(s[i]);
}

// The get_fwd_lgt functions return 1 if LGT is chosen to apply, and 0 otherwise
1082
int get_fwd_lgt4(transform_1d tx_orig, TxfmParam *txfm_param,
Lester Lu's avatar
Lester Lu committed
1083 1084 1085 1086
                 const tran_high_t *lgtmtx[], int ntx) {
  // inter/intra split
  if (tx_orig == &fadst4) {
    for (int i = 0; i < ntx; ++i)
1087
      lgtmtx[i] = txfm_param->is_inter ? &lgt4_170[0][0] : &lgt4_140[0][0];
Lester Lu's avatar
Lester Lu committed
1088 1089 1090 1091 1092
    return 1;
  }
  return 0;
}

1093
int get_fwd_lgt8(transform_1d tx_orig, TxfmParam *txfm_param,
Lester Lu's avatar
Lester Lu committed
1094 1095 1096 1097
                 const tran_high_t *lgtmtx[], int ntx) {
  // inter/intra split
  if (tx_orig == &fadst8) {
    for (int i = 0; i < ntx; ++i)
1098
      lgtmtx[i] = txfm_param->is_inter ? &lgt8_170[0][0] : &lgt8_150[0][0];
Lester Lu's avatar
Lester Lu committed
1099 1100 1101 1102 1103 1104
    return 1;
  }
  return 0;
}
#endif  // CONFIG_LGT

1105
#if CONFIG_EXT_TX
1106 1107 1108 1109
// TODO(sarahparker) these functions will be removed once the highbitdepth
// codepath works properly for rectangular transforms. They have almost
// identical versions in av1_fwd_txfm1d.c, but those are currently only
// being used for square transforms.
Debargha Mukherjee's avatar
Debargha Mukherjee committed
1110 1111
static void fidtx4(const tran_low_t *input, tran_low_t *output) {
  int i;
1112 1113 1114 1115
  for (i = 0; i < 4; ++i) {
#if CONFIG_DAALA_DCT4
    output[i] = input[i];
#else
Debargha Mukherjee's avatar
Debargha Mukherjee committed
1116
    output[i] = (tran_low_t)fdct_round_shift(input[i] * Sqrt2);
1117 1118
#endif
  }
Debargha Mukherjee's avatar
Debargha Mukherjee committed
1119 1120 1121 1122
}

static void fidtx8(const tran_low_t *input, tran_low_t *output) {
  int i;
1123
  for (i = 0; i < 8; ++i) output[i] = input[i] * 2;
Debargha Mukherjee's avatar
Debargha Mukherjee committed
1124 1125 1126 1127 1128 1129 1130 1131 1132 1133
}

static void fidtx16(const tran_low_t *input, tran_low_t *output) {
  int i;
  for (i = 0; i < 16; ++i)
    output[i] = (tran_low_t)fdct_round_shift(input[i] * 2 * Sqrt2);
}

static void fidtx32(const tran_low_t *input, tran_low_t *output) {
  int i;
1134
  for (i = 0; i < 32; ++i) output[i] = input[i] * 4;
Debargha Mukherjee's avatar
Debargha Mukherjee committed
1135 1136
}

1137
static void copy_block(const int16_t *src, int src_stride, int l, int w,
1138 1139 1140
                       int16_t *dest, int dest_stride) {
  int i;
  for (i = 0; i < l; ++i) {
1141
    memcpy(dest + dest_stride * i, src + src_stride * i, w * sizeof(int16_t));
1142 1143 1144
  }
}

1145
static void fliplr(int16_t *dest, int stride, int l, int w) {
1146 1147
  int i, j;
  for (i = 0; i < l; ++i) {
1148
    for (j = 0; j < w / 2; ++j) {
1149
      const int16_t tmp = dest[i * stride + j];
1150 1151
      dest[i * stride + j] = dest[i * stride + w - 1 - j];
      dest[i * stride + w - 1 - j] = tmp;
1152 1153 1154 1155
    }
  }
}

1156
static void flipud(int16_t *dest, int stride, int l, int w) {
1157
  int i, j;
1158
  for (j = 0; j < w; ++j) {
1159 1160 1161 1162 1163 1164 1165 1166
    for (i = 0; i < l / 2; ++i) {
      const int16_t tmp = dest[i * stride + j];
      dest[i * stride + j] = dest[(l - 1 - i) * stride + j];
      dest[(l - 1 - i) * stride + j] = tmp;
    }
  }
}

1167
static void fliplrud(int16_t *dest, int stride, int l, int w) {
1168 1169
  int i, j;
  for (i = 0; i < l / 2; ++i) {
1170
    for (j = 0; j < w; ++j) {
1171
      const int16_t tmp = dest[i * stride + j];
1172 1173
      dest[i * stride + j] = dest[(l - 1 - i) * stride + w - 1 - j];
      dest[(l - 1 - i) * stride + w - 1 - j] = tmp;
1174 1175 1176 1177
    }
  }
}

1178
static void copy_fliplr(const int16_t *src, int src_stride, int l, int w,
1179 1180 1181
                        int16_t *dest, int dest_stride) {
  copy_block(src, src_stride, l, w, dest, dest_stride);
  fliplr(dest, dest_stride, l, w);
1182 1183
}

1184
static void copy_flipud(const int16_t *src, int src_stride, int l, int w,
1185 1186 1187
                        int16_t *dest, int dest_stride) {
  copy_block(src, src_stride, l, w, dest, dest_stride);
  flipud(dest, dest_stride, l, w);
1188 1189
}

1190
static void copy_fliplrud(const int16_t *src, int src_stride, int l, int w,
1191 1192 1193
                          int16_t *dest, int dest_stride) {
  copy_block(src, src_stride, l, w, dest, dest_stride);
  fliplrud(dest, dest_stride, l, w);
1194 1195
}

1196
static void maybe_flip_input(const int16_t **src, int *src_stride, int l, int w,
1197 1198
                             int16_t *buff, int tx_type) {
  switch (tx_type) {
Sarah Parker's avatar
Sarah Parker committed
1199 1200 1201
#if CONFIG_MRC_TX
    case MRC_DCT:
#endif  // CONFIG_MRC_TX
1202 1203 1204 1205
    case DCT_DCT:
    case ADST_DCT:
    case DCT_ADST:
    case ADST_ADST:
Debargha Mukherjee's avatar
Debargha Mukherjee committed
1206
    case IDTX:
Jingning Han's avatar
Jingning Han committed
1207
    case V_DCT:
1208 1209
    case H_DCT:
    case V_ADST:
1210
    case H_ADST: break;
1211 1212
    case FLIPADST_DCT:
    case FLIPADST_ADST:
1213
    case V_FLIPADST:
1214
      copy_flipud(*src, *src_stride, l, w, buff, w);
1215
      *src = buff;
1216
      *src_stride = w;
1217 1218 1219
      break;
    case DCT_FLIPADST:
    case ADST_FLIPADST:
1220
    case H_FLIPADST:
1221
      copy_fliplr(*src, *src_stride, l, w, buff, w);
1222
      *src = buff;
1223
      *src_stride = w;
1224 1225
      break;
    case FLIPADST_FLIPADST:
1226
      copy_fliplrud(*src, *src_stride, l, w, buff, w);
1227
      *src = buff;
1228
      *src_stride = w;
1229
      break;
1230
    default: assert(0); break;
1231 1232 1233 1234
  }
}
#endif  // CONFIG_EXT_TX

Yaowu Xu's avatar
Yaowu Xu committed
1235
void av1_fht4x4_c(const int16_t *input, tran_low_t *output, int stride,
1236 1237
                  TxfmParam *txfm_param) {
  int tx_type = txfm_param->tx_type;
Sarah Parker's avatar
Sarah Parker committed
1238 1239 1240
#if CONFIG_MRC_TX
  assert(tx_type != MRC_DCT && "Invalid tx type for tx size");
#endif
1241 1242 1243
#if CONFIG_DCT_ONLY
  assert(tx_type == DCT_DCT);
#endif
1244
#if !CONFIG_DAALA_DCT4
Jingning Han's avatar
Jingning Han committed
1245
  if (tx_type == DCT_DCT) {
Yaowu Xu's avatar
Yaowu Xu committed
1246
    aom_fdct4x4_c(input, output, stride);
1247 1248 1249 1250
    return;
  }
#endif
  {
Geza Lore's avatar
Geza Lore committed
1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268
    static const transform_2d FHT[] = {
      { fdct4, fdct4 },    // DCT_DCT
      { fadst4, fdct4 },   // ADST_DCT
      { fdct4, fadst4 },   // DCT_ADST
      { fadst4, fadst4 },  // ADST_ADST
#if CONFIG_EXT_TX
      { fadst4, fdct4 },   // FLIPADST_DCT
      { fdct4, fadst4 },   // DCT_FLIPADST
      { fadst4, fadst4 },  // FLIPADST_FLIPADST
      { fadst4, fadst4 },  // ADST_FLIPADST
      { fadst4, fadst4 },  // FLIPADST_ADST
      { fidtx4, fidtx4 },  // IDTX
      { fdct4, fidtx4 },   // V_DCT
      { fidtx4, fdct4 },   // H_DCT
      { fadst4, fidtx4 },  // V_ADST
      { fidtx4, fadst4 },  // H_ADST
      { fadst4, fidtx4 },  // V_FLIPADST
      { fidtx4, fadst4 },  // H_FLIPADST
Lester Lu's avatar
Lester Lu committed
1269
#endif
Geza Lore's avatar
Geza Lore committed
1270 1271
    };
    const transform_2d ht = FHT[tx_type];
Jingning Han's avatar
Jingning Han committed
1272 1273 1274 1275
    tran_low_t out[4 * 4];
    int i, j;
    tran_low_t temp_in[4], temp_out[4];

1276 1277
#if CONFIG_EXT_TX
    int16_t flipped_input[4 * 4];
1278
    maybe_flip_input(&input, &stride, 4, 4, flipped_input, tx_type);
1279 1280
#endif

Lester Lu's avatar
Lester Lu committed
1281 1282 1283 1284 1285
#if CONFIG_LGT
    // Choose LGT adaptive to the prediction. We may apply different LGTs for
    // different rows/columns, indicated by the pointers to 2D arrays
    const tran_high_t *lgtmtx_col[4];
    const tran_high_t *lgtmtx_row[4];
1286 1287
    int use_lgt_col = get_fwd_lgt4(ht.cols, txfm_param, lgtmtx_col, 4);
    int use_lgt_row = get_fwd_lgt4(ht.rows, txfm_param, lgtmtx_row, 4);
Lester Lu's avatar
Lester Lu committed
1288 1289
#endif

Jingning Han's avatar
Jingning Han committed
1290 1291
    // Columns
    for (i = 0; i < 4; ++i) {
1292
      /* A C99-safe upshift by 4 for both Daala and VPx TX. */
1293
      for (j = 0; j < 4; ++j) temp_in[j] = input[j * stride + i] * 16;
1294
#if !CONFIG_DAALA_DCT4
1295
      if (i == 0 && temp_in[0]) temp_in[0] += 1;
1296
#endif
Lester Lu's avatar
Lester Lu committed
1297 1298 1299 1300 1301 1302
#if CONFIG_LGT
      if (use_lgt_col)
        flgt4(temp_in, temp_out, lgtmtx_col[i]);
      else
#endif
        ht.cols(temp_in, temp_out);
1303
      for (j = 0; j < 4; ++j) out[j * 4 + i] = temp_out[j];
Jingning Han's avatar
Jingning Han committed
1304 1305 1306 1307
    }

    // Rows
    for (i = 0; i < 4; ++i) {
1308
      for (j = 0; j < 4; ++j) temp_in[j] = out[j + i * 4];
Lester Lu's avatar
Lester Lu committed
1309 1310 1311 1312 1313 1314
#if CONFIG_LGT
      if (use_lgt_row)
        flgt4(temp_in, temp_out, lgtmtx_row[i]);
      else
#endif
        ht.rows(temp_in, temp_out);
1315 1316 1317 1318 1319
#if CONFIG_DAALA_DCT4
      /* Daala TX has orthonormal scaling; shift down by only 1 to achieve
         the usual VPx coefficient left-shift of 3. */
      for (j = 0; j < 4; ++j) output[j + i * 4] = temp_out[j] >> 1;
#else
1320
      for (j = 0; j < 4; ++j) output[j + i * 4] = (temp_out[j] + 1) >> 2;
1321
#endif
Jingning Han's avatar
Jingning Han committed
1322 1323 1324 1325
    }
  }
}

Yaowu Xu's avatar
Yaowu Xu committed
1326
void av1_fht4x8_c(const int16_t