idct.c 95.2 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
 */

#include <math.h>

Yaowu Xu's avatar
Yaowu Xu committed
14
#include "./aom_dsp_rtcd.h"
Geza Lore's avatar
Geza Lore committed
15 16 17 18
#include "./av1_rtcd.h"
#include "aom_dsp/inv_txfm.h"
#include "aom_ports/mem.h"
#include "av1/common/av1_inv_txfm2d_cfg.h"
19 20 21
#include "av1/common/blockd.h"
#include "av1/common/enums.h"
#include "av1/common/idct.h"
Jingning Han's avatar
Jingning Han committed
22

23
int get_tx_scale(const TX_SIZE tx_size) {
24 25 26 27 28 29 30
  if (txsize_sqr_up_map[tx_size] == TX_32X32) return 1;
#if CONFIG_TX64X64
  else if (txsize_sqr_up_map[tx_size] == TX_64X64)
    return 2;
#endif  // CONFIG_TX64X64
  else
    return 0;
31 32
}

33 34 35
// NOTE: The implementation of all inverses need to be aware of the fact
// that input and output could be the same buffer.

36
#if CONFIG_EXT_TX
Debargha Mukherjee's avatar
Debargha Mukherjee committed
37 38 39 40 41 42 43 44
static void iidtx4_c(const tran_low_t *input, tran_low_t *output) {
  int i;
  for (i = 0; i < 4; ++i)
    output[i] = (tran_low_t)dct_const_round_shift(input[i] * Sqrt2);
}

static void iidtx8_c(const tran_low_t *input, tran_low_t *output) {
  int i;
clang-format's avatar
clang-format committed
45
  for (i = 0; i < 8; ++i) output[i] = input[i] * 2;
Debargha Mukherjee's avatar
Debargha Mukherjee committed
46 47 48 49 50 51 52 53 54 55
}

static void iidtx16_c(const tran_low_t *input, tran_low_t *output) {
  int i;
  for (i = 0; i < 16; ++i)
    output[i] = (tran_low_t)dct_const_round_shift(input[i] * 2 * Sqrt2);
}

static void iidtx32_c(const tran_low_t *input, tran_low_t *output) {
  int i;
clang-format's avatar
clang-format committed
56
  for (i = 0; i < 32; ++i) output[i] = input[i] * 4;
Debargha Mukherjee's avatar
Debargha Mukherjee committed
57
}
58 59 60 61 62 63 64 65

#if CONFIG_TX64X64
static void iidtx64_c(const tran_low_t *input, tran_low_t *output) {
  int i;
  for (i = 0; i < 64; ++i)
    output[i] = (tran_low_t)dct_const_round_shift(input[i] * 4 * Sqrt2);
}
#endif  // CONFIG_TX64X64
66
#endif  // CONFIG_EXT_TX
Debargha Mukherjee's avatar
Debargha Mukherjee committed
67

68
// For use in lieu of ADST
69 70 71 72 73 74 75
static void ihalfright32_c(const tran_low_t *input, tran_low_t *output) {
  int i;
  tran_low_t inputhalf[16];
  // Multiply input by sqrt(2)
  for (i = 0; i < 16; ++i) {
    inputhalf[i] = (tran_low_t)dct_const_round_shift(input[i] * Sqrt2);
  }
76 77 78
  for (i = 0; i < 16; ++i) {
    output[i] = input[16 + i] * 4;
  }
Luca Barbato's avatar
Luca Barbato committed
79
  aom_idct16_c(inputhalf, output + 16);
80 81 82
  // Note overall scaling factor is 4 times orthogonal
}

83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117
#if CONFIG_TX64X64
static void idct64_col_c(const tran_low_t *input, tran_low_t *output) {
  int32_t in[64], out[64];
  int i;
  for (i = 0; i < 64; ++i) in[i] = (int32_t)input[i];
  av1_idct64_new(in, out, inv_cos_bit_col_dct_dct_64,
                 inv_stage_range_col_dct_dct_64);
  for (i = 0; i < 64; ++i) output[i] = (tran_low_t)out[i];
}

static void idct64_row_c(const tran_low_t *input, tran_low_t *output) {
  int32_t in[64], out[64];
  int i;
  for (i = 0; i < 64; ++i) in[i] = (int32_t)input[i];
  av1_idct64_new(in, out, inv_cos_bit_row_dct_dct_64,
                 inv_stage_range_row_dct_dct_64);
  for (i = 0; i < 64; ++i) output[i] = (tran_low_t)out[i];
}

// For use in lieu of ADST
static void ihalfright64_c(const tran_low_t *input, tran_low_t *output) {
  int i;
  tran_low_t inputhalf[32];
  // Multiply input by sqrt(2)
  for (i = 0; i < 32; ++i) {
    inputhalf[i] = (tran_low_t)dct_const_round_shift(input[i] * Sqrt2);
  }
  for (i = 0; i < 32; ++i) {
    output[i] = (tran_low_t)dct_const_round_shift(input[32 + i] * 4 * Sqrt2);
  }
  aom_idct32_c(inputhalf, output + 32);
  // Note overall scaling factor is 4 * sqrt(2)  times orthogonal
}
#endif  // CONFIG_TX64X64

118
#if CONFIG_HIGHBITDEPTH
119
#if CONFIG_EXT_TX
Debargha Mukherjee's avatar
Debargha Mukherjee committed
120 121 122 123
static void highbd_iidtx4_c(const tran_low_t *input, tran_low_t *output,
                            int bd) {
  int i;
  for (i = 0; i < 4; ++i)
124
    output[i] = HIGHBD_WRAPLOW(dct_const_round_shift(input[i] * Sqrt2), bd);
Debargha Mukherjee's avatar
Debargha Mukherjee committed
125 126 127 128 129
}

static void highbd_iidtx8_c(const tran_low_t *input, tran_low_t *output,
                            int bd) {
  int i;
clang-format's avatar
clang-format committed
130 131
  (void)bd;
  for (i = 0; i < 8; ++i) output[i] = input[i] * 2;
Debargha Mukherjee's avatar
Debargha Mukherjee committed
132 133 134
}

static void highbd_iidtx16_c(const tran_low_t *input, tran_low_t *output,
clang-format's avatar
clang-format committed
135
                             int bd) {
Debargha Mukherjee's avatar
Debargha Mukherjee committed
136 137
  int i;
  for (i = 0; i < 16; ++i)
138
    output[i] = HIGHBD_WRAPLOW(dct_const_round_shift(input[i] * 2 * Sqrt2), bd);
Debargha Mukherjee's avatar
Debargha Mukherjee committed
139 140 141 142 143
}

static void highbd_iidtx32_c(const tran_low_t *input, tran_low_t *output,
                             int bd) {
  int i;
clang-format's avatar
clang-format committed
144 145
  (void)bd;
  for (i = 0; i < 32; ++i) output[i] = input[i] * 4;
Debargha Mukherjee's avatar
Debargha Mukherjee committed
146
}
147
#endif  // CONFIG_EXT_TX
Debargha Mukherjee's avatar
Debargha Mukherjee committed
148

149 150 151 152 153 154
static void highbd_ihalfright32_c(const tran_low_t *input, tran_low_t *output,
                                  int bd) {
  int i;
  tran_low_t inputhalf[16];
  // Multiply input by sqrt(2)
  for (i = 0; i < 16; ++i) {
155
    inputhalf[i] = HIGHBD_WRAPLOW(dct_const_round_shift(input[i] * Sqrt2), bd);
156
  }
157 158 159
  for (i = 0; i < 16; ++i) {
    output[i] = input[16 + i] * 4;
  }
Yaowu Xu's avatar
Yaowu Xu committed
160
  aom_highbd_idct16_c(inputhalf, output + 16, bd);
161 162
  // Note overall scaling factor is 4 times orthogonal
}
163

164
#if CONFIG_EXT_TX
165 166 167 168 169
#if CONFIG_TX64X64
static void highbd_iidtx64_c(const tran_low_t *input, tran_low_t *output,
                             int bd) {
  int i;
  for (i = 0; i < 64; ++i)
170
    output[i] = HIGHBD_WRAPLOW(dct_const_round_shift(input[i] * 4 * Sqrt2), bd);
171
}
172 173
#endif  // CONFIG_TX64X64
#endif  // CONFIG_EXT_TX
174

175
#if CONFIG_TX64X64
176 177 178 179 180 181 182
// For use in lieu of ADST
static void highbd_ihalfright64_c(const tran_low_t *input, tran_low_t *output,
                                  int bd) {
  int i;
  tran_low_t inputhalf[32];
  // Multiply input by sqrt(2)
  for (i = 0; i < 32; ++i) {
183
    inputhalf[i] = HIGHBD_WRAPLOW(dct_const_round_shift(input[i] * Sqrt2), bd);
184 185
  }
  for (i = 0; i < 32; ++i) {
186 187
    output[i] =
        HIGHBD_WRAPLOW(dct_const_round_shift(input[32 + i] * 4 * Sqrt2), bd);
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
  }
  aom_highbd_idct32_c(inputhalf, output + 32, bd);
  // Note overall scaling factor is 4 * sqrt(2)  times orthogonal
}

static void highbd_idct64_col_c(const tran_low_t *input, tran_low_t *output,
                                int bd) {
  int32_t in[64], out[64];
  int i;
  (void)bd;
  for (i = 0; i < 64; ++i) in[i] = (int32_t)input[i];
  av1_idct64_new(in, out, inv_cos_bit_col_dct_dct_64,
                 inv_stage_range_col_dct_dct_64);
  for (i = 0; i < 64; ++i) output[i] = (tran_low_t)out[i];
}

static void highbd_idct64_row_c(const tran_low_t *input, tran_low_t *output,
                                int bd) {
  int32_t in[64], out[64];
  int i;
  (void)bd;
  for (i = 0; i < 64; ++i) in[i] = (int32_t)input[i];
  av1_idct64_new(in, out, inv_cos_bit_row_dct_dct_64,
                 inv_stage_range_row_dct_dct_64);
  for (i = 0; i < 64; ++i) output[i] = (tran_low_t)out[i];
}
#endif  // CONFIG_TX64X64
215
#endif  // CONFIG_HIGHBITDEPTH
216

Jingning Han's avatar
Jingning Han committed
217
// Inverse identity transform and add.
218
#if CONFIG_EXT_TX
219
static void inv_idtx_add_c(const tran_low_t *input, uint8_t *dest, int stride,
Jingning Han's avatar
Jingning Han committed
220
                           int bs, int tx_type) {
221
  int r, c;
222
  const int shift = bs < 32 ? 3 : (bs < 64 ? 2 : 1);
Debargha Mukherjee's avatar
Debargha Mukherjee committed
223
  if (tx_type == IDTX) {
Jingning Han's avatar
Jingning Han committed
224 225
    for (r = 0; r < bs; ++r) {
      for (c = 0; c < bs; ++c)
Debargha Mukherjee's avatar
Debargha Mukherjee committed
226 227 228
        dest[c] = clip_pixel_add(dest[c], input[c] >> shift);
      dest += stride;
      input += bs;
Jingning Han's avatar
Jingning Han committed
229
    }
230 231
  }
}
232
#endif  // CONFIG_EXT_TX
233

clang-format's avatar
clang-format committed
234 235 236 237 238
#define FLIPUD_PTR(dest, stride, size)       \
  do {                                       \
    (dest) = (dest) + ((size)-1) * (stride); \
    (stride) = -(stride);                    \
  } while (0)
239

240
#if CONFIG_EXT_TX
clang-format's avatar
clang-format committed
241 242 243
static void maybe_flip_strides(uint8_t **dst, int *dstride, tran_low_t **src,
                               int *sstride, int tx_type, int sizey,
                               int sizex) {
244 245 246 247 248 249 250 251
  // Note that the transpose of src will be added to dst. In order to LR
  // flip the addends (in dst coordinates), we UD flip the src. To UD flip
  // the addends, we UD flip the dst.
  switch (tx_type) {
    case DCT_DCT:
    case ADST_DCT:
    case DCT_ADST:
    case ADST_ADST:
Debargha Mukherjee's avatar
Debargha Mukherjee committed
252
    case IDTX:
Jingning Han's avatar
Jingning Han committed
253 254
    case V_DCT:
    case H_DCT:
255
    case V_ADST:
clang-format's avatar
clang-format committed
256
    case H_ADST: break;
257 258
    case FLIPADST_DCT:
    case FLIPADST_ADST:
259
    case V_FLIPADST:
260
      // flip UD
261
      FLIPUD_PTR(*dst, *dstride, sizey);
262 263 264
      break;
    case DCT_FLIPADST:
    case ADST_FLIPADST:
265
    case H_FLIPADST:
266
      // flip LR
267
      FLIPUD_PTR(*src, *sstride, sizex);
268 269 270
      break;
    case FLIPADST_FLIPADST:
      // flip UD
271
      FLIPUD_PTR(*dst, *dstride, sizey);
272
      // flip LR
273
      FLIPUD_PTR(*src, *sstride, sizex);
274
      break;
clang-format's avatar
clang-format committed
275
    default: assert(0); break;
276 277
  }
}
278
#endif  // CONFIG_EXT_TX
279

280
#if CONFIG_HIGHBITDEPTH
281
#if CONFIG_EXT_TX
282
static void highbd_inv_idtx_add_c(const tran_low_t *input, uint8_t *dest8,
283
                                  int stride, int bs, int tx_type, int bd) {
284 285 286
  int r, c;
  const int shift = bs < 32 ? 3 : 2;
  uint16_t *dest = CONVERT_TO_SHORTPTR(dest8);
287

Debargha Mukherjee's avatar
Debargha Mukherjee committed
288
  if (tx_type == IDTX) {
289 290
    for (r = 0; r < bs; ++r) {
      for (c = 0; c < bs; ++c)
Debargha Mukherjee's avatar
Debargha Mukherjee committed
291 292 293
        dest[c] = highbd_clip_pixel_add(dest[c], input[c] >> shift, bd);
      dest += stride;
      input += bs;
294
    }
295 296
  }
}
297

clang-format's avatar
clang-format committed
298 299 300
static void maybe_flip_strides16(uint16_t **dst, int *dstride, tran_low_t **src,
                                 int *sstride, int tx_type, int sizey,
                                 int sizex) {
301 302 303 304 305 306 307 308
  // Note that the transpose of src will be added to dst. In order to LR
  // flip the addends (in dst coordinates), we UD flip the src. To UD flip
  // the addends, we UD flip the dst.
  switch (tx_type) {
    case DCT_DCT:
    case ADST_DCT:
    case DCT_ADST:
    case ADST_ADST:
Debargha Mukherjee's avatar
Debargha Mukherjee committed
309 310 311
    case IDTX:
    case V_DCT:
    case H_DCT:
312
    case V_ADST:
clang-format's avatar
clang-format committed
313
    case H_ADST: break;
314 315
    case FLIPADST_DCT:
    case FLIPADST_ADST:
316
    case V_FLIPADST:
317
      // flip UD
318
      FLIPUD_PTR(*dst, *dstride, sizey);
319 320 321
      break;
    case DCT_FLIPADST:
    case ADST_FLIPADST:
322
    case H_FLIPADST:
323
      // flip LR
324
      FLIPUD_PTR(*src, *sstride, sizex);
325 326 327
      break;
    case FLIPADST_FLIPADST:
      // flip UD
328
      FLIPUD_PTR(*dst, *dstride, sizey);
329
      // flip LR
330
      FLIPUD_PTR(*src, *sstride, sizex);
331
      break;
clang-format's avatar
clang-format committed
332
    default: assert(0); break;
333 334
  }
}
335
#endif  // CONFIG_EXT_TX
336
#endif  // CONFIG_HIGHBITDEPTH
337

Yaowu Xu's avatar
Yaowu Xu committed
338 339
void av1_iht4x4_16_add_c(const tran_low_t *input, uint8_t *dest, int stride,
                         int tx_type) {
340
  static const transform_2d IHT_4[] = {
Luca Barbato's avatar
Luca Barbato committed
341 342 343 344
    { aom_idct4_c, aom_idct4_c },    // DCT_DCT  = 0
    { aom_iadst4_c, aom_idct4_c },   // ADST_DCT = 1
    { aom_idct4_c, aom_iadst4_c },   // DCT_ADST = 2
    { aom_iadst4_c, aom_iadst4_c },  // ADST_ADST = 3
345
#if CONFIG_EXT_TX
Luca Barbato's avatar
Luca Barbato committed
346 347 348 349 350 351 352 353 354 355 356 357 358
    { aom_iadst4_c, aom_idct4_c },   // FLIPADST_DCT
    { aom_idct4_c, aom_iadst4_c },   // DCT_FLIPADST
    { aom_iadst4_c, aom_iadst4_c },  // FLIPADST_FLIPADST
    { aom_iadst4_c, aom_iadst4_c },  // ADST_FLIPADST
    { aom_iadst4_c, aom_iadst4_c },  // FLIPADST_ADST
    { iidtx4_c, iidtx4_c },          // IDTX
    { aom_idct4_c, iidtx4_c },       // V_DCT
    { iidtx4_c, aom_idct4_c },       // H_DCT
    { aom_iadst4_c, iidtx4_c },      // V_ADST
    { iidtx4_c, aom_iadst4_c },      // H_ADST
    { aom_iadst4_c, iidtx4_c },      // V_FLIPADST
    { iidtx4_c, aom_iadst4_c },      // H_FLIPADST
#endif                               // CONFIG_EXT_TX
359 360 361
  };

  int i, j;
362 363 364 365
  tran_low_t tmp;
  tran_low_t out[4][4];
  tran_low_t *outp = &out[0][0];
  int outstride = 4;
366 367 368

  // inverse transform row vectors
  for (i = 0; i < 4; ++i) {
369
    IHT_4[tx_type].rows(input, out[i]);
clang-format's avatar
clang-format committed
370
    input += 4;
371 372 373
  }

  // transpose
clang-format's avatar
clang-format committed
374
  for (i = 1; i < 4; i++) {
375
    for (j = 0; j < i; j++) {
clang-format's avatar
clang-format committed
376
      tmp = out[i][j];
377 378 379
      out[i][j] = out[j][i];
      out[j][i] = tmp;
    }
380 381 382 383
  }

  // inverse transform column vectors
  for (i = 0; i < 4; ++i) {
384 385 386 387
    IHT_4[tx_type].cols(out[i], out[i]);
  }

#if CONFIG_EXT_TX
388
  maybe_flip_strides(&dest, &stride, &outp, &outstride, tx_type, 4, 4);
389 390 391 392
#endif

  // Sum with the destination
  for (i = 0; i < 4; ++i) {
393
    for (j = 0; j < 4; ++j) {
394 395 396
      int d = i * stride + j;
      int s = j * outstride + i;
      dest[d] = clip_pixel_add(dest[d], ROUND_POWER_OF_TWO(outp[s], 4));
397 398 399 400
    }
  }
}

Yaowu Xu's avatar
Yaowu Xu committed
401 402
void av1_iht4x8_32_add_c(const tran_low_t *input, uint8_t *dest, int stride,
                         int tx_type) {
403
  static const transform_2d IHT_4x8[] = {
Luca Barbato's avatar
Luca Barbato committed
404 405 406 407
    { aom_idct8_c, aom_idct4_c },    // DCT_DCT
    { aom_iadst8_c, aom_idct4_c },   // ADST_DCT
    { aom_idct8_c, aom_iadst4_c },   // DCT_ADST
    { aom_iadst8_c, aom_iadst4_c },  // ADST_ADST
408
#if CONFIG_EXT_TX
Luca Barbato's avatar
Luca Barbato committed
409 410 411 412 413 414 415 416 417 418 419 420
    { aom_iadst8_c, aom_idct4_c },   // FLIPADST_DCT
    { aom_idct8_c, aom_iadst4_c },   // DCT_FLIPADST
    { aom_iadst8_c, aom_iadst4_c },  // FLIPADST_FLIPADST
    { aom_iadst8_c, aom_iadst4_c },  // ADST_FLIPADST
    { aom_iadst8_c, aom_iadst4_c },  // FLIPADST_ADST
    { iidtx8_c, iidtx4_c },          // IDTX
    { aom_idct8_c, iidtx4_c },       // V_DCT
    { iidtx8_c, aom_idct4_c },       // H_DCT
    { aom_iadst8_c, iidtx4_c },      // V_ADST
    { iidtx8_c, aom_iadst4_c },      // H_ADST
    { aom_iadst8_c, iidtx4_c },      // V_FLIPADST
    { iidtx8_c, aom_iadst4_c },      // H_FLIPADST
421
#endif
422 423
  };

424 425
  const int n = 4;
  const int n2 = 8;
426 427 428
  int i, j;
  tran_low_t out[4][8], outtmp[4];
  tran_low_t *outp = &out[0][0];
429
  int outstride = n2;
430 431

  // inverse transform row vectors and transpose
432
  for (i = 0; i < n2; ++i) {
433
    IHT_4x8[tx_type].rows(input, outtmp);
434
    for (j = 0; j < n; ++j)
435
      out[j][i] = (tran_low_t)dct_const_round_shift(outtmp[j] * Sqrt2);
436
    input += n;
437 438 439
  }

  // inverse transform column vectors
440
  for (i = 0; i < n; ++i) {
441 442 443
    IHT_4x8[tx_type].cols(out[i], out[i]);
  }

444
#if CONFIG_EXT_TX
445
  maybe_flip_strides(&dest, &stride, &outp, &outstride, tx_type, n2, n);
446
#endif
447 448

  // Sum with the destination
449 450
  for (i = 0; i < n2; ++i) {
    for (j = 0; j < n; ++j) {
451 452 453 454 455 456 457
      int d = i * stride + j;
      int s = j * outstride + i;
      dest[d] = clip_pixel_add(dest[d], ROUND_POWER_OF_TWO(outp[s], 5));
    }
  }
}

Yaowu Xu's avatar
Yaowu Xu committed
458 459
void av1_iht8x4_32_add_c(const tran_low_t *input, uint8_t *dest, int stride,
                         int tx_type) {
460
  static const transform_2d IHT_8x4[] = {
Luca Barbato's avatar
Luca Barbato committed
461 462 463 464
    { aom_idct4_c, aom_idct8_c },    // DCT_DCT
    { aom_iadst4_c, aom_idct8_c },   // ADST_DCT
    { aom_idct4_c, aom_iadst8_c },   // DCT_ADST
    { aom_iadst4_c, aom_iadst8_c },  // ADST_ADST
465
#if CONFIG_EXT_TX
Luca Barbato's avatar
Luca Barbato committed
466 467 468 469 470 471 472 473 474 475 476 477
    { aom_iadst4_c, aom_idct8_c },   // FLIPADST_DCT
    { aom_idct4_c, aom_iadst8_c },   // DCT_FLIPADST
    { aom_iadst4_c, aom_iadst8_c },  // FLIPADST_FLIPADST
    { aom_iadst4_c, aom_iadst8_c },  // ADST_FLIPADST
    { aom_iadst4_c, aom_iadst8_c },  // FLIPADST_ADST
    { iidtx4_c, iidtx8_c },          // IDTX
    { aom_idct4_c, iidtx8_c },       // V_DCT
    { iidtx4_c, aom_idct8_c },       // H_DCT
    { aom_iadst4_c, iidtx8_c },      // V_ADST
    { iidtx4_c, aom_iadst8_c },      // H_ADST
    { aom_iadst4_c, iidtx8_c },      // V_FLIPADST
    { iidtx4_c, aom_iadst8_c },      // H_FLIPADST
478
#endif
479
  };
480 481
  const int n = 4;
  const int n2 = 8;
482 483 484 485

  int i, j;
  tran_low_t out[8][4], outtmp[8];
  tran_low_t *outp = &out[0][0];
486
  int outstride = n;
487 488

  // inverse transform row vectors and transpose
489
  for (i = 0; i < n; ++i) {
490
    IHT_8x4[tx_type].rows(input, outtmp);
491
    for (j = 0; j < n2; ++j)
492
      out[j][i] = (tran_low_t)dct_const_round_shift(outtmp[j] * Sqrt2);
493
    input += n2;
494 495 496
  }

  // inverse transform column vectors
497
  for (i = 0; i < n2; ++i) {
498 499 500
    IHT_8x4[tx_type].cols(out[i], out[i]);
  }

501
#if CONFIG_EXT_TX
502
  maybe_flip_strides(&dest, &stride, &outp, &outstride, tx_type, n, n2);
503
#endif
504 505

  // Sum with the destination
506 507
  for (i = 0; i < n; ++i) {
    for (j = 0; j < n2; ++j) {
508 509 510 511 512 513 514
      int d = i * stride + j;
      int s = j * outstride + i;
      dest[d] = clip_pixel_add(dest[d], ROUND_POWER_OF_TWO(outp[s], 5));
    }
  }
}

515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622
void av1_iht4x16_64_add_c(const tran_low_t *input, uint8_t *dest, int stride,
                          int tx_type) {
  static const transform_2d IHT_4x16[] = {
    { aom_idct16_c, aom_idct4_c },    // DCT_DCT
    { aom_iadst16_c, aom_idct4_c },   // ADST_DCT
    { aom_idct16_c, aom_iadst4_c },   // DCT_ADST
    { aom_iadst16_c, aom_iadst4_c },  // ADST_ADST
#if CONFIG_EXT_TX
    { aom_iadst16_c, aom_idct4_c },   // FLIPADST_DCT
    { aom_idct16_c, aom_iadst4_c },   // DCT_FLIPADST
    { aom_iadst16_c, aom_iadst4_c },  // FLIPADST_FLIPADST
    { aom_iadst16_c, aom_iadst4_c },  // ADST_FLIPADST
    { aom_iadst16_c, aom_iadst4_c },  // FLIPADST_ADST
    { iidtx16_c, iidtx4_c },          // IDTX
    { aom_idct16_c, iidtx4_c },       // V_DCT
    { iidtx16_c, aom_idct4_c },       // H_DCT
    { aom_iadst16_c, iidtx4_c },      // V_ADST
    { iidtx16_c, aom_iadst4_c },      // H_ADST
    { aom_iadst16_c, iidtx4_c },      // V_FLIPADST
    { iidtx16_c, aom_iadst4_c },      // H_FLIPADST
#endif
  };

  const int n = 4;
  const int n4 = 16;
  int i, j;
  tran_low_t out[4][16], outtmp[4];
  tran_low_t *outp = &out[0][0];
  int outstride = n4;

  // inverse transform row vectors and transpose
  for (i = 0; i < n4; ++i) {
    IHT_4x16[tx_type].rows(input, outtmp);
    for (j = 0; j < n; ++j) out[j][i] = outtmp[j];
    input += n;
  }

  // inverse transform column vectors
  for (i = 0; i < n; ++i) IHT_4x16[tx_type].cols(out[i], out[i]);

#if CONFIG_EXT_TX
  maybe_flip_strides(&dest, &stride, &outp, &outstride, tx_type, n4, n);
#endif

  // Sum with the destination
  for (i = 0; i < n4; ++i) {
    for (j = 0; j < n; ++j) {
      int d = i * stride + j;
      int s = j * outstride + i;
      dest[d] = clip_pixel_add(dest[d], ROUND_POWER_OF_TWO(outp[s], 5));
    }
  }
}

void av1_iht16x4_64_add_c(const tran_low_t *input, uint8_t *dest, int stride,
                          int tx_type) {
  static const transform_2d IHT_16x4[] = {
    { aom_idct4_c, aom_idct16_c },    // DCT_DCT
    { aom_iadst4_c, aom_idct16_c },   // ADST_DCT
    { aom_idct4_c, aom_iadst16_c },   // DCT_ADST
    { aom_iadst4_c, aom_iadst16_c },  // ADST_ADST
#if CONFIG_EXT_TX
    { aom_iadst4_c, aom_idct16_c },   // FLIPADST_DCT
    { aom_idct4_c, aom_iadst16_c },   // DCT_FLIPADST
    { aom_iadst4_c, aom_iadst16_c },  // FLIPADST_FLIPADST
    { aom_iadst4_c, aom_iadst16_c },  // ADST_FLIPADST
    { aom_iadst4_c, aom_iadst16_c },  // FLIPADST_ADST
    { iidtx4_c, iidtx16_c },          // IDTX
    { aom_idct4_c, iidtx16_c },       // V_DCT
    { iidtx4_c, aom_idct16_c },       // H_DCT
    { aom_iadst4_c, iidtx16_c },      // V_ADST
    { iidtx4_c, aom_iadst16_c },      // H_ADST
    { aom_iadst4_c, iidtx16_c },      // V_FLIPADST
    { iidtx4_c, aom_iadst16_c },      // H_FLIPADST
#endif
  };
  const int n = 4;
  const int n4 = 16;

  int i, j;
  tran_low_t out[16][4], outtmp[16];
  tran_low_t *outp = &out[0][0];
  int outstride = n;

  // inverse transform row vectors and transpose
  for (i = 0; i < n; ++i) {
    IHT_16x4[tx_type].rows(input, outtmp);
    for (j = 0; j < n4; ++j) out[j][i] = outtmp[j];
    input += n4;
  }

  // inverse transform column vectors
  for (i = 0; i < n4; ++i) IHT_16x4[tx_type].cols(out[i], out[i]);

#if CONFIG_EXT_TX
  maybe_flip_strides(&dest, &stride, &outp, &outstride, tx_type, n, n4);
#endif

  // Sum with the destination
  for (i = 0; i < n; ++i) {
    for (j = 0; j < n4; ++j) {
      int d = i * stride + j;
      int s = j * outstride + i;
      dest[d] = clip_pixel_add(dest[d], ROUND_POWER_OF_TWO(outp[s], 5));
    }
  }
}

Yaowu Xu's avatar
Yaowu Xu committed
623 624
void av1_iht8x16_128_add_c(const tran_low_t *input, uint8_t *dest, int stride,
                           int tx_type) {
625
  static const transform_2d IHT_8x16[] = {
Luca Barbato's avatar
Luca Barbato committed
626 627 628 629
    { aom_idct16_c, aom_idct8_c },    // DCT_DCT
    { aom_iadst16_c, aom_idct8_c },   // ADST_DCT
    { aom_idct16_c, aom_iadst8_c },   // DCT_ADST
    { aom_iadst16_c, aom_iadst8_c },  // ADST_ADST
630
#if CONFIG_EXT_TX
Luca Barbato's avatar
Luca Barbato committed
631 632 633 634 635 636 637 638 639 640 641 642
    { aom_iadst16_c, aom_idct8_c },   // FLIPADST_DCT
    { aom_idct16_c, aom_iadst8_c },   // DCT_FLIPADST
    { aom_iadst16_c, aom_iadst8_c },  // FLIPADST_FLIPADST
    { aom_iadst16_c, aom_iadst8_c },  // ADST_FLIPADST
    { aom_iadst16_c, aom_iadst8_c },  // FLIPADST_ADST
    { iidtx16_c, iidtx8_c },          // IDTX
    { aom_idct16_c, iidtx8_c },       // V_DCT
    { iidtx16_c, aom_idct8_c },       // H_DCT
    { aom_iadst16_c, iidtx8_c },      // V_ADST
    { iidtx16_c, aom_iadst8_c },      // H_ADST
    { aom_iadst16_c, iidtx8_c },      // V_FLIPADST
    { iidtx16_c, aom_iadst8_c },      // H_FLIPADST
643
#endif
644 645 646 647 648 649 650 651 652 653 654 655 656 657
  };

  const int n = 8;
  const int n2 = 16;
  int i, j;
  tran_low_t out[8][16], outtmp[8];
  tran_low_t *outp = &out[0][0];
  int outstride = n2;

  // inverse transform row vectors and transpose
  for (i = 0; i < n2; ++i) {
    IHT_8x16[tx_type].rows(input, outtmp);
    for (j = 0; j < n; ++j)
      out[j][i] = (tran_low_t)dct_const_round_shift(outtmp[j] * Sqrt2);
clang-format's avatar
clang-format committed
658
    input += n;
659 660 661 662 663 664 665
  }

  // inverse transform column vectors
  for (i = 0; i < n; ++i) {
    IHT_8x16[tx_type].cols(out[i], out[i]);
  }

666
#if CONFIG_EXT_TX
667
  maybe_flip_strides(&dest, &stride, &outp, &outstride, tx_type, n2, n);
668
#endif
669 670 671 672 673 674 675 676 677 678 679

  // Sum with the destination
  for (i = 0; i < n2; ++i) {
    for (j = 0; j < n; ++j) {
      int d = i * stride + j;
      int s = j * outstride + i;
      dest[d] = clip_pixel_add(dest[d], ROUND_POWER_OF_TWO(outp[s], 6));
    }
  }
}

Yaowu Xu's avatar
Yaowu Xu committed
680 681
void av1_iht16x8_128_add_c(const tran_low_t *input, uint8_t *dest, int stride,
                           int tx_type) {
682
  static const transform_2d IHT_16x8[] = {
Luca Barbato's avatar
Luca Barbato committed
683 684 685 686
    { aom_idct8_c, aom_idct16_c },    // DCT_DCT
    { aom_iadst8_c, aom_idct16_c },   // ADST_DCT
    { aom_idct8_c, aom_iadst16_c },   // DCT_ADST
    { aom_iadst8_c, aom_iadst16_c },  // ADST_ADST
687
#if CONFIG_EXT_TX
Luca Barbato's avatar
Luca Barbato committed
688 689 690 691 692 693 694 695 696 697 698 699
    { aom_iadst8_c, aom_idct16_c },   // FLIPADST_DCT
    { aom_idct8_c, aom_iadst16_c },   // DCT_FLIPADST
    { aom_iadst8_c, aom_iadst16_c },  // FLIPADST_FLIPADST
    { aom_iadst8_c, aom_iadst16_c },  // ADST_FLIPADST
    { aom_iadst8_c, aom_iadst16_c },  // FLIPADST_ADST
    { iidtx8_c, iidtx16_c },          // IDTX
    { aom_idct8_c, iidtx16_c },       // V_DCT
    { iidtx8_c, aom_idct16_c },       // H_DCT
    { aom_iadst8_c, iidtx16_c },      // V_ADST
    { iidtx8_c, aom_iadst16_c },      // H_ADST
    { aom_iadst8_c, iidtx16_c },      // V_FLIPADST
    { iidtx8_c, aom_iadst16_c },      // H_FLIPADST
700
#endif
701 702 703 704 705 706 707 708 709 710 711 712 713 714
  };
  const int n = 8;
  const int n2 = 16;

  int i, j;
  tran_low_t out[16][8], outtmp[16];
  tran_low_t *outp = &out[0][0];
  int outstride = n;

  // inverse transform row vectors and transpose
  for (i = 0; i < n; ++i) {
    IHT_16x8[tx_type].rows(input, outtmp);
    for (j = 0; j < n2; ++j)
      out[j][i] = (tran_low_t)dct_const_round_shift(outtmp[j] * Sqrt2);
clang-format's avatar
clang-format committed
715
    input += n2;
716 717 718 719 720 721 722
  }

  // inverse transform column vectors
  for (i = 0; i < n2; ++i) {
    IHT_16x8[tx_type].cols(out[i], out[i]);
  }

723
#if CONFIG_EXT_TX
724
  maybe_flip_strides(&dest, &stride, &outp, &outstride, tx_type, n, n2);
725
#endif
726 727 728 729 730 731 732 733 734 735 736

  // Sum with the destination
  for (i = 0; i < n; ++i) {
    for (j = 0; j < n2; ++j) {
      int d = i * stride + j;
      int s = j * outstride + i;
      dest[d] = clip_pixel_add(dest[d], ROUND_POWER_OF_TWO(outp[s], 6));
    }
  }
}

737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844
void av1_iht8x32_256_add_c(const tran_low_t *input, uint8_t *dest, int stride,
                           int tx_type) {
  static const transform_2d IHT_8x32[] = {
    { aom_idct32_c, aom_idct8_c },     // DCT_DCT
    { ihalfright32_c, aom_idct8_c },   // ADST_DCT
    { aom_idct32_c, aom_iadst8_c },    // DCT_ADST
    { ihalfright32_c, aom_iadst8_c },  // ADST_ADST
#if CONFIG_EXT_TX
    { ihalfright32_c, aom_idct8_c },   // FLIPADST_DCT
    { aom_idct32_c, aom_iadst8_c },    // DCT_FLIPADST
    { ihalfright32_c, aom_iadst8_c },  // FLIPADST_FLIPADST
    { ihalfright32_c, aom_iadst8_c },  // ADST_FLIPADST
    { ihalfright32_c, aom_iadst8_c },  // FLIPADST_ADST
    { iidtx32_c, iidtx8_c },           // IDTX
    { aom_idct32_c, iidtx8_c },        // V_DCT
    { iidtx32_c, aom_idct8_c },        // H_DCT
    { ihalfright32_c, iidtx8_c },      // V_ADST
    { iidtx32_c, aom_iadst8_c },       // H_ADST
    { ihalfright32_c, iidtx8_c },      // V_FLIPADST
    { iidtx32_c, aom_iadst8_c },       // H_FLIPADST
#endif
  };

  const int n = 8;
  const int n4 = 32;
  int i, j;
  tran_low_t out[8][32], outtmp[8];
  tran_low_t *outp = &out[0][0];
  int outstride = n4;

  // inverse transform row vectors and transpose
  for (i = 0; i < n4; ++i) {
    IHT_8x32[tx_type].rows(input, outtmp);
    for (j = 0; j < n; ++j) out[j][i] = outtmp[j];
    input += n;
  }

  // inverse transform column vectors
  for (i = 0; i < n; ++i) IHT_8x32[tx_type].cols(out[i], out[i]);

#if CONFIG_EXT_TX
  maybe_flip_strides(&dest, &stride, &outp, &outstride, tx_type, n4, n);
#endif

  // Sum with the destination
  for (i = 0; i < n4; ++i) {
    for (j = 0; j < n; ++j) {
      int d = i * stride + j;
      int s = j * outstride + i;
      dest[d] = clip_pixel_add(dest[d], ROUND_POWER_OF_TWO(outp[s], 6));
    }
  }
}

void av1_iht32x8_256_add_c(const tran_low_t *input, uint8_t *dest, int stride,
                           int tx_type) {
  static const transform_2d IHT_32x8[] = {
    { aom_idct8_c, aom_idct32_c },     // DCT_DCT
    { aom_iadst8_c, aom_idct32_c },    // ADST_DCT
    { aom_idct8_c, ihalfright32_c },   // DCT_ADST
    { aom_iadst8_c, ihalfright32_c },  // ADST_ADST
#if CONFIG_EXT_TX
    { aom_iadst8_c, aom_idct32_c },    // FLIPADST_DCT
    { aom_idct8_c, ihalfright32_c },   // DCT_FLIPADST
    { aom_iadst8_c, ihalfright32_c },  // FLIPADST_FLIPADST
    { aom_iadst8_c, ihalfright32_c },  // ADST_FLIPADST
    { aom_iadst8_c, ihalfright32_c },  // FLIPADST_ADST
    { iidtx8_c, iidtx32_c },           // IDTX
    { aom_idct8_c, iidtx32_c },        // V_DCT
    { iidtx8_c, aom_idct32_c },        // H_DCT
    { aom_iadst8_c, iidtx32_c },       // V_ADST
    { iidtx8_c, ihalfright32_c },      // H_ADST
    { aom_iadst8_c, iidtx32_c },       // V_FLIPADST
    { iidtx8_c, ihalfright32_c },      // H_FLIPADST
#endif
  };
  const int n = 8;
  const int n4 = 32;

  int i, j;
  tran_low_t out[32][8], outtmp[32];
  tran_low_t *outp = &out[0][0];
  int outstride = n;

  // inverse transform row vectors and transpose
  for (i = 0; i < n; ++i) {
    IHT_32x8[tx_type].rows(input, outtmp);
    for (j = 0; j < n4; ++j) out[j][i] = outtmp[j];
    input += n4;
  }

  // inverse transform column vectors
  for (i = 0; i < n4; ++i) IHT_32x8[tx_type].cols(out[i], out[i]);

#if CONFIG_EXT_TX
  maybe_flip_strides(&dest, &stride, &outp, &outstride, tx_type, n, n4);
#endif

  // Sum with the destination
  for (i = 0; i < n; ++i) {
    for (j = 0; j < n4; ++j) {
      int d = i * stride + j;
      int s = j * outstride + i;
      dest[d] = clip_pixel_add(dest[d], ROUND_POWER_OF_TWO(outp[s], 6));
    }
  }
}

Yaowu Xu's avatar
Yaowu Xu committed
845 846
void av1_iht16x32_512_add_c(const tran_low_t *input, uint8_t *dest, int stride,
                            int tx_type) {
847
  static const transform_2d IHT_16x32[] = {
Luca Barbato's avatar
Luca Barbato committed
848 849 850 851
    { aom_idct32_c, aom_idct16_c },     // DCT_DCT
    { ihalfright32_c, aom_idct16_c },   // ADST_DCT
    { aom_idct32_c, aom_iadst16_c },    // DCT_ADST
    { ihalfright32_c, aom_iadst16_c },  // ADST_ADST
852
#if CONFIG_EXT_TX
Luca Barbato's avatar
Luca Barbato committed
853 854 855 856 857 858 859 860 861 862 863 864
    { ihalfright32_c, aom_idct16_c },   // FLIPADST_DCT
    { aom_idct32_c, aom_iadst16_c },    // DCT_FLIPADST
    { ihalfright32_c, aom_iadst16_c },  // FLIPADST_FLIPADST
    { ihalfright32_c, aom_iadst16_c },  // ADST_FLIPADST
    { ihalfright32_c, aom_iadst16_c },  // FLIPADST_ADST
    { iidtx32_c, iidtx16_c },           // IDTX
    { aom_idct32_c, iidtx16_c },        // V_DCT
    { iidtx32_c, aom_idct16_c },        // H_DCT
    { ihalfright32_c, iidtx16_c },      // V_ADST
    { iidtx32_c, aom_iadst16_c },       // H_ADST
    { ihalfright32_c, iidtx16_c },      // V_FLIPADST
    { iidtx32_c, aom_iadst16_c },       // H_FLIPADST
865
#endif
866 867 868 869 870 871 872 873 874 875 876 877 878 879
  };

  const int n = 16;
  const int n2 = 32;
  int i, j;
  tran_low_t out[16][32], outtmp[16];
  tran_low_t *outp = &out[0][0];
  int outstride = n2;

  // inverse transform row vectors and transpose
  for (i = 0; i < n2; ++i) {
    IHT_16x32[tx_type].rows(input, outtmp);
    for (j = 0; j < n; ++j)
      out[j][i] = (tran_low_t)dct_const_round_shift(outtmp[j] * Sqrt2);
clang-format's avatar
clang-format committed
880
    input += n;
881 882 883 884 885 886 887
  }

  // inverse transform column vectors
  for (i = 0; i < n; ++i) {
    IHT_16x32[tx_type].cols(out[i], out[i]);
  }

888
#if CONFIG_EXT_TX
889
  maybe_flip_strides(&dest, &stride, &outp, &outstride, tx_type, n2, n);
890
#endif
891 892 893 894 895 896 897 898 899 900 901

  // Sum with the destination
  for (i = 0; i < n2; ++i) {
    for (j = 0; j < n; ++j) {
      int d = i * stride + j;
      int s = j * outstride + i;
      dest[d] = clip_pixel_add(dest[d], ROUND_POWER_OF_TWO(outp[s], 6));
    }
  }
}

Yaowu Xu's avatar
Yaowu Xu committed
902 903
void av1_iht32x16_512_add_c(const tran_low_t *input, uint8_t *dest, int stride,
                            int tx_type) {
904
  static const transform_2d IHT_32x16[] = {
Luca Barbato's avatar
Luca Barbato committed
905 906 907 908
    { aom_idct16_c, aom_idct32_c },     // DCT_DCT
    { aom_iadst16_c, aom_idct32_c },    // ADST_DCT
    { aom_idct16_c, ihalfright32_c },   // DCT_ADST
    { aom_iadst16_c, ihalfright32_c },  // ADST_ADST
909
#if CONFIG_EXT_TX
Luca Barbato's avatar
Luca Barbato committed
910 911 912 913 914 915 916 917 918 919 920 921
    { aom_iadst16_c, aom_idct32_c },    // FLIPADST_DCT
    { aom_idct16_c, ihalfright32_c },   // DCT_FLIPADST
    { aom_iadst16_c, ihalfright32_c },  // FLIPADST_FLIPADST
    { aom_iadst16_c, ihalfright32_c },  // ADST_FLIPADST
    { aom_iadst16_c, ihalfright32_c },  // FLIPADST_ADST
    { iidtx16_c, iidtx32_c },           // IDTX
    { aom_idct16_c, iidtx32_c },        // V_DCT
    { iidtx16_c, aom_idct32_c },        // H_DCT
    { aom_iadst16_c, iidtx32_c },       // V_ADST
    { iidtx16_c, ihalfright32_c },      // H_ADST
    { aom_iadst16_c, iidtx32_c },       // V_FLIPADST
    { iidtx16_c, ihalfright32_c },      // H_FLIPADST
922
#endif
923 924 925 926 927 928 929 930 931 932 933 934 935 936
  };
  const int n = 16;
  const int n2 = 32;

  int i, j;
  tran_low_t out[32][16], outtmp[32];
  tran_low_t *outp = &out[0][0];
  int outstride = n;

  // inverse transform row vectors and transpose
  for (i = 0; i < n; ++i) {
    IHT_32x16[tx_type].rows(input, outtmp);
    for (j = 0; j < n2; ++j)
      out[j][i] = (tran_low_t)dct_const_round_shift(outtmp[j] * Sqrt2);
clang-format's avatar
clang-format committed
937
    input += n2;
938 939 940 941 942 943 944
  }

  // inverse transform column vectors
  for (i = 0; i < n2; ++i) {
    IHT_32x16[tx_type].cols(out[i], out[i]);
  }

945
#if CONFIG_EXT_TX
946
  maybe_flip_strides(&dest, &stride, &outp, &outstride, tx_type, n, n2);
947
#endif
948 949 950 951 952 953 954 955 956 957 958

  // Sum with the destination
  for (i = 0; i < n; ++i) {
    for (j = 0; j < n2; ++j) {
      int d = i * stride + j;
      int s = j * outstride + i;
      dest[d] = clip_pixel_add(dest[d], ROUND_POWER_OF_TWO(outp[s], 6));
    }
  }
}

Yaowu Xu's avatar
Yaowu Xu committed
959 960
void av1_iht8x8_64_add_c(const tran_low_t *input, uint8_t *dest, int stride,
                         int tx_type) {
961
  static const transform_2d IHT_8[] = {
Luca Barbato's avatar
Luca Barbato committed
962 963 964 965
    { aom_idct8_c, aom_idct8_c },    // DCT_DCT  = 0
    { aom_iadst8_c, aom_idct8_c },   // ADST_DCT = 1
    { aom_idct8_c, aom_iadst8_c },   // DCT_ADST = 2
    { aom_iadst8_c, aom_iadst8_c },  // ADST_ADST = 3
966
#if CONFIG_EXT_TX
Luca Barbato's avatar
Luca Barbato committed
967 968 969 970 971 972 973 974 975 976 977 978 979
    { aom_iadst8_c, aom_idct8_c },   // FLIPADST_DCT
    { aom_idct8_c, aom_iadst8_c },   // DCT_FLIPADST
    { aom_iadst8_c, aom_iadst8_c },  // FLIPADST_FLIPADST
    { aom_iadst8_c, aom_iadst8_c },  // ADST_FLIPADST
    { aom_iadst8_c, aom_iadst8_c },  // FLIPADST_ADST
    { iidtx8_c, iidtx8_c },          // IDTX
    { aom_idct8_c, iidtx8_c },       // V_DCT
    { iidtx8_c, aom_idct8_c },       // H_DCT
    { aom_iadst8_c, iidtx8_c },      // V_ADST
    { iidtx8_c, aom_iadst8_c },      // H_ADST
    { aom_iadst8_c, iidtx8_c },      // V_FLIPADST
    { iidtx8_c, aom_iadst8_c },      // H_FLIPADST
#endif                               // CONFIG_EXT_TX
980 981
  };

Jingning Han's avatar
Jingning Han committed
982
  int i, j;
983 984 985 986
  tran_low_t tmp;
  tran_low_t out[8][8];
  tran_low_t *outp = &out[0][0];
  int outstride = 8;
Jingning Han's avatar
Jingning Han committed
987 988 989

  // inverse transform row vectors
  for (i = 0; i < 8; ++i) {
990
    IHT_8[tx_type].rows(input, out[i]);
clang-format's avatar
clang-format committed
991
    input += 8;
992 993 994
  }

  // transpose
clang-format's avatar
clang-format committed
995
  for (i = 1; i < 8; i++) {
996
    for (j = 0; j < i; j++) {
clang-format's avatar
clang-format committed
997
      tmp = out[i][j];
998 999 1000
      out[i][j] = out[j][i];
      out[j][i] = tmp;
    }
Jingning Han's avatar
Jingning Han committed
1001 1002 1003 1004
  }

  // inverse transform column vectors
  for (i = 0; i < 8; ++i) {
1005 1006 1007 1008
    IHT_8[tx_type].cols(out[i], out[i]);
  }

#if CONFIG_EXT_TX
1009
  maybe_flip_strides(&dest, &stride, &outp, &outstride, tx_type, 8, 8);
1010 1011 1012 1013
#endif

  // Sum with the destination
  for (i = 0; i < 8; ++i) {
Jingning Han's avatar
Jingning Han committed
1014
    for (j = 0; j < 8; ++j) {
1015 1016 1017
      int d = i * stride + j;
      int s = j * outstride + i;
      dest[d] = clip_pixel_add(dest[d], ROUND_POWER_OF_TWO(outp[s], 5));
Jingning Han's avatar
Jingning Han committed
1018 1019 1020 1021
    }
  }
}

Yaowu Xu's avatar
Yaowu Xu committed
1022 1023
void av1_iht16x16_256_add_c(const tran_low_t *input, uint8_t *dest, int stride,
                            int tx_type) {
1024
  static const transform_2d IHT_16[] = {
Luca Barbato's avatar
Luca Barbato committed
1025 1026 1027 1028
    { aom_idct16_c, aom_idct16_c },    // DCT_DCT  = 0
    { aom_iadst16_c, aom_idct16_c },   // ADST_DCT = 1
    { aom_idct16_c, aom_iadst16_c },   // DCT_ADST = 2
    { aom_iadst16_c, aom_iadst16_c },  // ADST_ADST = 3
1029
#if CONFIG_EXT_TX
Luca Barbato's avatar
Luca Barbato committed
1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042
    { aom_iadst16_c, aom_idct16_c },   // FLIPADST_DCT
    { aom_idct16_c, aom_iadst16_c },   // DCT_FLIPADST
    { aom_iadst16_c, aom_iadst16_c },  // FLIPADST_FLIPADST
    { aom_iadst16_c, aom_iadst16_c },  // ADST_FLIPADST
    { aom_iadst16_c, aom_iadst16_c },  // FLIPADST_ADST
    { iidtx16_c, iidtx16_c },          // IDTX
    { aom_idct16_c, iidtx16_c },       // V_DCT
    { iidtx16_c, aom_idct16_c },       // H_DCT
    { aom_iadst16_c, iidtx16_c },      // V_ADST
    { iidtx16_c, aom_iadst16_c },      // H_ADST
    { aom_iadst16_c, iidtx16_c },      // V_FLIPADST
    { iidtx16_c, aom_iadst16_c },      // H_FLIPADST
#endif                                 // CONFIG_EXT_TX
1043
  };
1044

Jingning Han's avatar
Jingning Han committed
1045
  int i, j;
1046 1047 1048 1049 1050 1051 1052 1053
  tran_low_t tmp;
  tran_low_t out[16][16];
  tran_low_t *outp = &out[0][0];
  int outstride = 16;

  // inverse transform row vectors
  for (i = 0; i < 16; ++i) {
    IHT_16[tx_type].rows(input, out[i]);
clang-format's avatar
clang-format committed
1054
    input += 16;
1055 1056 1057
  }

  // transpose
clang-format's avatar
clang-format committed
1058
  for (i = 1; i < 16; i++) {
1059
    for (j = 0; j < i; j++) {
clang-format's avatar
clang-format committed
1060
      tmp = out[i][j];