encodemb.c 64.3 KB
Newer Older
Jingning Han's avatar
Jingning Han committed
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
/*
 *  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
 *
 *  Use of this source code is governed by a BSD-style license
 *  that can be found in the LICENSE file in the root of the source
 *  tree. An additional intellectual property rights grant can be found
 *  in the file PATENTS.  All contributing project authors may
 *  be found in the AUTHORS file in the root of the source tree.
 */


#include "./vp10_rtcd.h"
#include "./vpx_config.h"
#include "./vpx_dsp_rtcd.h"

#include "vpx_dsp/quantize.h"
#include "vpx_mem/vpx_mem.h"
#include "vpx_ports/mem.h"

20
21
22
23
#include "vp10/common/idct.h"
#include "vp10/common/reconinter.h"
#include "vp10/common/reconintra.h"
#include "vp10/common/scan.h"
Jingning Han's avatar
Jingning Han committed
24

25
26
27
#include "vp10/encoder/encodemb.h"
#include "vp10/encoder/rd.h"
#include "vp10/encoder/tokenize.h"
Jingning Han's avatar
Jingning Han committed
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106

struct optimize_ctx {
  ENTROPY_CONTEXT ta[MAX_MB_PLANE][16];
  ENTROPY_CONTEXT tl[MAX_MB_PLANE][16];
};

void vp10_subtract_plane(MACROBLOCK *x, BLOCK_SIZE bsize, int plane) {
  struct macroblock_plane *const p = &x->plane[plane];
  const struct macroblockd_plane *const pd = &x->e_mbd.plane[plane];
  const BLOCK_SIZE plane_bsize = get_plane_block_size(bsize, pd);
  const int bw = 4 * num_4x4_blocks_wide_lookup[plane_bsize];
  const int bh = 4 * num_4x4_blocks_high_lookup[plane_bsize];

#if CONFIG_VP9_HIGHBITDEPTH
  if (x->e_mbd.cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
    vpx_highbd_subtract_block(bh, bw, p->src_diff, bw, p->src.buf,
                              p->src.stride, pd->dst.buf, pd->dst.stride,
                              x->e_mbd.bd);
    return;
  }
#endif  // CONFIG_VP9_HIGHBITDEPTH
  vpx_subtract_block(bh, bw, p->src_diff, bw, p->src.buf, p->src.stride,
                     pd->dst.buf, pd->dst.stride);
}

#define RDTRUNC(RM, DM, R, D) ((128 + (R) * (RM)) & 0xFF)

typedef struct vp10_token_state {
  int           rate;
  int           error;
  int           next;
  int16_t       token;
  short         qc;
} vp10_token_state;

// TODO(jimbankoski): experiment to find optimal RD numbers.
static const int plane_rd_mult[PLANE_TYPES] = { 4, 2 };

#define UPDATE_RD_COST()\
{\
  rd_cost0 = RDCOST(rdmult, rddiv, rate0, error0);\
  rd_cost1 = RDCOST(rdmult, rddiv, rate1, error1);\
  if (rd_cost0 == rd_cost1) {\
    rd_cost0 = RDTRUNC(rdmult, rddiv, rate0, error0);\
    rd_cost1 = RDTRUNC(rdmult, rddiv, rate1, error1);\
  }\
}

// This function is a place holder for now but may ultimately need
// to scan previous tokens to work out the correct context.
static int trellis_get_coeff_context(const int16_t *scan,
                                     const int16_t *nb,
                                     int idx, int token,
                                     uint8_t *token_cache) {
  int bak = token_cache[scan[idx]], pt;
  token_cache[scan[idx]] = vp10_pt_energy_class[token];
  pt = get_coef_context(nb, token_cache, idx + 1);
  token_cache[scan[idx]] = bak;
  return pt;
}

static int optimize_b(MACROBLOCK *mb, int plane, int block,
                      TX_SIZE tx_size, int ctx) {
  MACROBLOCKD *const xd = &mb->e_mbd;
  struct macroblock_plane *const p = &mb->plane[plane];
  struct macroblockd_plane *const pd = &xd->plane[plane];
  const int ref = is_inter_block(&xd->mi[0]->mbmi);
  vp10_token_state tokens[1025][2];
  unsigned best_index[1025][2];
  uint8_t token_cache[1024];
  const tran_low_t *const coeff = BLOCK_OFFSET(mb->plane[plane].coeff, block);
  tran_low_t *const qcoeff = BLOCK_OFFSET(p->qcoeff, block);
  tran_low_t *const dqcoeff = BLOCK_OFFSET(pd->dqcoeff, block);
  const int eob = p->eobs[block];
  const PLANE_TYPE type = pd->plane_type;
  const int default_eob = 16 << (tx_size << 1);
  const int mul = 1 + (tx_size == TX_32X32);
  const int16_t *dequant_ptr = pd->dequant;
  const uint8_t *const band_translate = get_band_translate(tx_size);
hui su's avatar
hui su committed
107
  TX_TYPE tx_type = get_tx_type(type, xd, block, tx_size);
108
109
  const scan_order *const so =
      get_scan(tx_size, tx_type, is_inter_block(&xd->mi[0]->mbmi));
Jingning Han's avatar
Jingning Han committed
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
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
  const int16_t *const scan = so->scan;
  const int16_t *const nb = so->neighbors;
  int next = eob, sz = 0;
  int64_t rdmult = mb->rdmult * plane_rd_mult[type], rddiv = mb->rddiv;
  int64_t rd_cost0, rd_cost1;
  int rate0, rate1, error0, error1;
  int16_t t0, t1;
  EXTRABIT e0;
  int best, band, pt, i, final_eob;
#if CONFIG_VP9_HIGHBITDEPTH
  const int16_t *cat6_high_cost = vp10_get_high_cost_table(xd->bd);
#else
  const int16_t *cat6_high_cost = vp10_get_high_cost_table(8);
#endif

  assert((!type && !plane) || (type && plane));
  assert(eob <= default_eob);

  /* Now set up a Viterbi trellis to evaluate alternative roundings. */
  if (!ref)
    rdmult = (rdmult * 9) >> 4;

  /* Initialize the sentinel node of the trellis. */
  tokens[eob][0].rate = 0;
  tokens[eob][0].error = 0;
  tokens[eob][0].next = default_eob;
  tokens[eob][0].token = EOB_TOKEN;
  tokens[eob][0].qc = 0;
  tokens[eob][1] = tokens[eob][0];

  for (i = 0; i < eob; i++)
    token_cache[scan[i]] =
        vp10_pt_energy_class[vp10_get_token(qcoeff[scan[i]])];

  for (i = eob; i-- > 0;) {
    int base_bits, d2, dx;
    const int rc = scan[i];
    int x = qcoeff[rc];
    /* Only add a trellis state for non-zero coefficients. */
    if (x) {
      int shortcut = 0;
      error0 = tokens[next][0].error;
      error1 = tokens[next][1].error;
      /* Evaluate the first possibility for this state. */
      rate0 = tokens[next][0].rate;
      rate1 = tokens[next][1].rate;
      vp10_get_token_extra(x, &t0, &e0);
      /* Consider both possible successor states. */
      if (next < default_eob) {
        band = band_translate[i + 1];
        pt = trellis_get_coeff_context(scan, nb, i, t0, token_cache);
        rate0 += mb->token_costs[tx_size][type][ref][band][0][pt]
                                [tokens[next][0].token];
        rate1 += mb->token_costs[tx_size][type][ref][band][0][pt]
                                [tokens[next][1].token];
      }
      UPDATE_RD_COST();
      /* And pick the best. */
      best = rd_cost1 < rd_cost0;
      base_bits = vp10_get_cost(t0, e0, cat6_high_cost);
      dx = mul * (dqcoeff[rc] - coeff[rc]);
#if CONFIG_VP9_HIGHBITDEPTH
      if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
        dx >>= xd->bd - 8;
      }
#endif  // CONFIG_VP9_HIGHBITDEPTH
      d2 = dx * dx;
      tokens[i][0].rate = base_bits + (best ? rate1 : rate0);
      tokens[i][0].error = d2 + (best ? error1 : error0);
      tokens[i][0].next = next;
      tokens[i][0].token = t0;
      tokens[i][0].qc = x;
      best_index[i][0] = best;

      /* Evaluate the second possibility for this state. */
      rate0 = tokens[next][0].rate;
      rate1 = tokens[next][1].rate;

      if ((abs(x) * dequant_ptr[rc != 0] > abs(coeff[rc]) * mul) &&
          (abs(x) * dequant_ptr[rc != 0] < abs(coeff[rc]) * mul +
                                               dequant_ptr[rc != 0]))
        shortcut = 1;
      else
        shortcut = 0;

      if (shortcut) {
        sz = -(x < 0);
        x -= 2 * sz + 1;
      }

      /* Consider both possible successor states. */
      if (!x) {
        /* If we reduced this coefficient to zero, check to see if
         *  we need to move the EOB back here.
         */
        t0 = tokens[next][0].token == EOB_TOKEN ? EOB_TOKEN : ZERO_TOKEN;
        t1 = tokens[next][1].token == EOB_TOKEN ? EOB_TOKEN : ZERO_TOKEN;
        e0 = 0;
      } else {
        vp10_get_token_extra(x, &t0, &e0);
        t1 = t0;
      }
      if (next < default_eob) {
        band = band_translate[i + 1];
        if (t0 != EOB_TOKEN) {
          pt = trellis_get_coeff_context(scan, nb, i, t0, token_cache);
          rate0 += mb->token_costs[tx_size][type][ref][band][!x][pt]
                                  [tokens[next][0].token];
        }
        if (t1 != EOB_TOKEN) {
          pt = trellis_get_coeff_context(scan, nb, i, t1, token_cache);
          rate1 += mb->token_costs[tx_size][type][ref][band][!x][pt]
                                  [tokens[next][1].token];
        }
      }

      UPDATE_RD_COST();
      /* And pick the best. */
      best = rd_cost1 < rd_cost0;
      base_bits = vp10_get_cost(t0, e0, cat6_high_cost);

      if (shortcut) {
#if CONFIG_VP9_HIGHBITDEPTH
        if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
          dx -= ((dequant_ptr[rc != 0] >> (xd->bd - 8)) + sz) ^ sz;
        } else {
          dx -= (dequant_ptr[rc != 0] + sz) ^ sz;
        }
#else
        dx -= (dequant_ptr[rc != 0] + sz) ^ sz;
#endif  // CONFIG_VP9_HIGHBITDEPTH
        d2 = dx * dx;
      }
      tokens[i][1].rate = base_bits + (best ? rate1 : rate0);
      tokens[i][1].error = d2 + (best ? error1 : error0);
      tokens[i][1].next = next;
      tokens[i][1].token = best ? t1 : t0;
      tokens[i][1].qc = x;
      best_index[i][1] = best;
      /* Finally, make this the new head of the trellis. */
      next = i;
    } else {
      /* There's no choice to make for a zero coefficient, so we don't
       *  add a new trellis node, but we do need to update the costs.
       */
      band = band_translate[i + 1];
      t0 = tokens[next][0].token;
      t1 = tokens[next][1].token;
      /* Update the cost of each path if we're past the EOB token. */
      if (t0 != EOB_TOKEN) {
        tokens[next][0].rate +=
            mb->token_costs[tx_size][type][ref][band][1][0][t0];
        tokens[next][0].token = ZERO_TOKEN;
      }
      if (t1 != EOB_TOKEN) {
        tokens[next][1].rate +=
            mb->token_costs[tx_size][type][ref][band][1][0][t1];
        tokens[next][1].token = ZERO_TOKEN;
      }
      best_index[i][0] = best_index[i][1] = 0;
      /* Don't update next, because we didn't add a new node. */
    }
  }

  /* Now pick the best path through the whole trellis. */
  band = band_translate[i + 1];
  rate0 = tokens[next][0].rate;
  rate1 = tokens[next][1].rate;
  error0 = tokens[next][0].error;
  error1 = tokens[next][1].error;
  t0 = tokens[next][0].token;
  t1 = tokens[next][1].token;
  rate0 += mb->token_costs[tx_size][type][ref][band][0][ctx][t0];
  rate1 += mb->token_costs[tx_size][type][ref][band][0][ctx][t1];
  UPDATE_RD_COST();
  best = rd_cost1 < rd_cost0;
  final_eob = -1;
  memset(qcoeff, 0, sizeof(*qcoeff) * (16 << (tx_size * 2)));
  memset(dqcoeff, 0, sizeof(*dqcoeff) * (16 << (tx_size * 2)));
  for (i = next; i < eob; i = next) {
    const int x = tokens[i][best].qc;
    const int rc = scan[i];
    if (x) {
      final_eob = i;
    }

    qcoeff[rc] = x;
    dqcoeff[rc] = (x * dequant_ptr[rc != 0]) / mul;

    next = tokens[i][best].next;
    best = best_index[i][best];
  }
  final_eob++;

  mb->plane[plane].eobs[block] = final_eob;
305
  assert(final_eob <= default_eob);
Jingning Han's avatar
Jingning Han committed
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
  return final_eob;
}

static INLINE void fdct32x32(int rd_transform,
                             const int16_t *src, tran_low_t *dst,
                             int src_stride) {
  if (rd_transform)
    vpx_fdct32x32_rd(src, dst, src_stride);
  else
    vpx_fdct32x32(src, dst, src_stride);
}

#if CONFIG_VP9_HIGHBITDEPTH
static INLINE void highbd_fdct32x32(int rd_transform, const int16_t *src,
                                    tran_low_t *dst, int src_stride) {
  if (rd_transform)
    vpx_highbd_fdct32x32_rd(src, dst, src_stride);
  else
    vpx_highbd_fdct32x32(src, dst, src_stride);
}
#endif  // CONFIG_VP9_HIGHBITDEPTH

328
329
330
331
332
333
334
335
336
#if CONFIG_EXT_TX
static void copy_block(const int16_t *src, int src_stride, int l,
                       int16_t *dest, int dest_stride) {
  int i;
  for (i = 0; i < l; ++i) {
    memcpy(dest + dest_stride * i, src + src_stride * i,
           l * sizeof(int16_t));
  }
}
Jingning Han's avatar
Jingning Han committed
337

338
339
340
341
342
343
344
static void fliplr(int16_t *dest, int stride, int l) {
  int i, j;
  for (i = 0; i < l; ++i) {
    for (j = 0; j < l / 2; ++j) {
      const int16_t tmp = dest[i * stride + j];
      dest[i * stride + j] = dest[i * stride + l - 1 - j];
      dest[i * stride + l - 1 - j] = tmp;
Jingning Han's avatar
Jingning Han committed
345
346
    }
  }
347
}
Jingning Han's avatar
Jingning Han committed
348

349
350
351
352
353
354
355
356
static void flipud(int16_t *dest, int stride, int l) {
  int i, j;
  for (j = 0; j < l; ++j) {
    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;
    }
Jingning Han's avatar
Jingning Han committed
357
358
359
  }
}

360
static void fliplrud(int16_t *dest, int stride, int l) {
Jingning Han's avatar
Jingning Han committed
361
  int i, j;
362
363
364
365
366
  for (i = 0; i < l / 2; ++i) {
    for (j = 0; j < l; ++j) {
      const int16_t tmp = dest[i * stride + j];
      dest[i * stride + j] = dest[(l - 1 - i) * stride + l - 1 - j];
      dest[(l - 1 - i) * stride + l - 1 - j] = tmp;
Jingning Han's avatar
Jingning Han committed
367
368
    }
  }
369
}
Jingning Han's avatar
Jingning Han committed
370

371
372
373
374
375
376
377
378
379
380
static void copy_fliplr(const int16_t *src, int src_stride, int l,
                          int16_t *dest, int dest_stride) {
  copy_block(src, src_stride, l, dest, dest_stride);
  fliplr(dest, dest_stride, l);
}

static void copy_flipud(const int16_t *src, int src_stride, int l,
                          int16_t *dest, int dest_stride) {
  copy_block(src, src_stride, l, dest, dest_stride);
  flipud(dest, dest_stride, l);
Jingning Han's avatar
Jingning Han committed
381
382
}

383
384
385
386
387
static void copy_fliplrud(const int16_t *src, int src_stride, int l,
                            int16_t *dest, int dest_stride) {
  copy_block(src, src_stride, l, dest, dest_stride);
  fliplrud(dest, dest_stride, l);
}
388
389
390
391
392
393
394
395
396
397
398
399
400
401

// Forward identity transform.
static void fwd_idtx_c(const int16_t *src_diff, tran_low_t *coeff, int stride,
                       int bs) {
  int r, c;
  const int shift = bs < 32 ? 3 : 2;

  for (r = 0; r < bs; ++r) {
    for (c = 0; c < bs; ++c)
      coeff[c] = src_diff[c] << shift;
    src_diff += stride;
    coeff += bs;
  }
}
402
403
#endif  // CONFIG_EXT_TX

404
405
406
407
408
void vp10_fwd_txfm_4x4(const int16_t *src_diff, tran_low_t *coeff,
                       int diff_stride, TX_TYPE tx_type, int lossless) {
  if (lossless) {
    vp10_fwht4x4(src_diff, coeff, diff_stride);
  } else {
409
410
411
#if CONFIG_EXT_TX
  int16_t src_diff2[16];
#endif  // CONFIG_EXT_TX
hui su's avatar
hui su committed
412
413
  switch (tx_type) {
    case DCT_DCT:
414
      vpx_fdct4x4(src_diff, coeff, diff_stride);
hui su's avatar
hui su committed
415
416
417
418
419
420
      break;
    case ADST_DCT:
    case DCT_ADST:
    case ADST_ADST:
      vp10_fht4x4(src_diff, coeff, diff_stride, tx_type);
      break;
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
#if CONFIG_EXT_TX
    case FLIPADST_DCT:
      copy_flipud(src_diff, diff_stride, 4, src_diff2, 4);
      vp10_fht4x4(src_diff2, coeff, 4, ADST_DCT);
      break;
    case DCT_FLIPADST:
      copy_fliplr(src_diff, diff_stride, 4, src_diff2, 4);
      vp10_fht4x4(src_diff2, coeff, 4, DCT_ADST);
      break;
    case FLIPADST_FLIPADST:
      copy_fliplrud(src_diff, diff_stride, 4, src_diff2, 4);
      vp10_fht4x4(src_diff2, coeff, 4, ADST_ADST);
      break;
    case ADST_FLIPADST:
      copy_fliplr(src_diff, diff_stride, 4, src_diff2, 4);
      vp10_fht4x4(src_diff2, coeff, 4, ADST_ADST);
      break;
    case FLIPADST_ADST:
      copy_flipud(src_diff, diff_stride, 4, src_diff2, 4);
      vp10_fht4x4(src_diff2, coeff, 4, ADST_ADST);
      break;
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
    case DST_DST:
    case DCT_DST:
    case DST_DCT:
    case DST_ADST:
    case ADST_DST:
      // Use C version since DST exists only in C
      vp10_fht4x4_c(src_diff, coeff, diff_stride, tx_type);
      break;
    case DST_FLIPADST:
      copy_fliplr(src_diff, diff_stride, 4, src_diff2, 4);
      vp10_fht4x4_c(src_diff2, coeff, 4, DST_ADST);
      break;
    case FLIPADST_DST:
      copy_flipud(src_diff, diff_stride, 4, src_diff2, 4);
      vp10_fht4x4_c(src_diff2, coeff, 4, ADST_DST);
      break;
458
459
460
    case IDTX:
      fwd_idtx_c(src_diff, coeff, diff_stride, 4);
      break;
461
#endif  // CONFIG_EXT_TX
hui su's avatar
hui su committed
462
463
464
    default:
      assert(0);
      break;
465
    }
hui su's avatar
hui su committed
466
467
468
469
470
  }
}

static void fwd_txfm_8x8(const int16_t *src_diff, tran_low_t *coeff,
                         int diff_stride, TX_TYPE tx_type) {
471
472
473
#if CONFIG_EXT_TX
  int16_t src_diff2[64];
#endif  // CONFIG_EXT_TX
hui su's avatar
hui su committed
474
475
476
477
478
479
480
  switch (tx_type) {
    case DCT_DCT:
    case ADST_DCT:
    case DCT_ADST:
    case ADST_ADST:
      vp10_fht8x8(src_diff, coeff, diff_stride, tx_type);
      break;
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
#if CONFIG_EXT_TX
    case FLIPADST_DCT:
      copy_flipud(src_diff, diff_stride, 8, src_diff2, 8);
      vp10_fht8x8(src_diff2, coeff, 8, ADST_DCT);
      break;
    case DCT_FLIPADST:
      copy_fliplr(src_diff, diff_stride, 8, src_diff2, 8);
      vp10_fht8x8(src_diff2, coeff, 8, DCT_ADST);
      break;
    case FLIPADST_FLIPADST:
      copy_fliplrud(src_diff, diff_stride, 8, src_diff2, 8);
      vp10_fht8x8(src_diff2, coeff, 8, ADST_ADST);
      break;
    case ADST_FLIPADST:
      copy_fliplr(src_diff, diff_stride, 8, src_diff2, 8);
      vp10_fht8x8(src_diff2, coeff, 8, ADST_ADST);
      break;
    case FLIPADST_ADST:
      copy_flipud(src_diff, diff_stride, 8, src_diff2, 8);
      vp10_fht8x8(src_diff2, coeff, 8, ADST_ADST);
      break;
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
    case DST_DST:
    case DCT_DST:
    case DST_DCT:
    case DST_ADST:
    case ADST_DST:
      // Use C version since DST exists only in C
      vp10_fht8x8_c(src_diff, coeff, diff_stride, tx_type);
      break;
    case DST_FLIPADST:
      copy_fliplr(src_diff, diff_stride, 8, src_diff2, 8);
      vp10_fht8x8_c(src_diff2, coeff, 8, DST_ADST);
      break;
    case FLIPADST_DST:
      copy_flipud(src_diff, diff_stride, 8, src_diff2, 8);
      vp10_fht8x8_c(src_diff2, coeff, 8, ADST_DST);
      break;
518
519
520
    case IDTX:
      fwd_idtx_c(src_diff, coeff, diff_stride, 8);
      break;
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
#endif  // CONFIG_EXT_TX
    default:
      assert(0);
      break;
  }
}

static void fwd_txfm_8x8_1(const int16_t *src_diff, tran_low_t *coeff,
                           int diff_stride, TX_TYPE tx_type) {
#if CONFIG_EXT_TX
  int16_t src_diff2[64];
#endif  // CONFIG_EXT_TX
  switch (tx_type) {
    case DCT_DCT:
    case ADST_DCT:
    case DCT_ADST:
    case ADST_ADST:
      vpx_fdct8x8_1(src_diff, coeff, diff_stride);
      break;
#if CONFIG_EXT_TX
    case FLIPADST_DCT:
      copy_flipud(src_diff, diff_stride, 8, src_diff2, 8);
      vp10_fht8x8(src_diff2, coeff, 8, ADST_DCT);
      break;
    case DCT_FLIPADST:
      copy_fliplr(src_diff, diff_stride, 8, src_diff2, 8);
      vp10_fht8x8(src_diff2, coeff, 8, DCT_ADST);
      break;
    case FLIPADST_FLIPADST:
      copy_fliplrud(src_diff, diff_stride, 8, src_diff2, 8);
      vp10_fht8x8(src_diff2, coeff, 8, ADST_ADST);
      break;
    case ADST_FLIPADST:
      copy_fliplr(src_diff, diff_stride, 8, src_diff2, 8);
      vp10_fht8x8(src_diff2, coeff, 8, ADST_ADST);
      break;
    case FLIPADST_ADST:
      copy_flipud(src_diff, diff_stride, 8, src_diff2, 8);
      vp10_fht8x8(src_diff2, coeff, 8, ADST_ADST);
      break;
561
562
563
564
565
566
567
    case DST_DST:
    case DCT_DST:
    case DST_DCT:
    case DST_ADST:
    case ADST_DST:
      // Use C version since DST exists only in C
      vp10_fht8x8_c(src_diff, coeff, diff_stride, tx_type);
568
        break;
569
570
571
572
573
574
575
576
    case DST_FLIPADST:
      copy_fliplr(src_diff, diff_stride, 8, src_diff2, 8);
      vp10_fht8x8_c(src_diff2, coeff, 8, DST_ADST);
      break;
    case FLIPADST_DST:
      copy_flipud(src_diff, diff_stride, 8, src_diff2, 8);
      vp10_fht8x8_c(src_diff2, coeff, 8, ADST_DST);
      break;
577
578
579
    case IDTX:
      fwd_idtx_c(src_diff, coeff, diff_stride, 8);
      break;
580
#endif  // CONFIG_EXT_TX
581
582
583
      default:
        assert(0);
        break;
hui su's avatar
hui su committed
584
585
586
587
588
  }
}

static void fwd_txfm_16x16(const int16_t *src_diff, tran_low_t *coeff,
                           int diff_stride, TX_TYPE tx_type) {
589
590
591
#if CONFIG_EXT_TX
  int16_t src_diff2[256];
#endif  // CONFIG_EXT_TX
hui su's avatar
hui su committed
592
593
594
595
596
597
598
  switch (tx_type) {
    case DCT_DCT:
    case ADST_DCT:
    case DCT_ADST:
    case ADST_ADST:
      vp10_fht16x16(src_diff, coeff, diff_stride, tx_type);
      break;
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
#if CONFIG_EXT_TX
    case FLIPADST_DCT:
      copy_flipud(src_diff, diff_stride, 16, src_diff2, 16);
      vp10_fht16x16(src_diff2, coeff, 16, ADST_DCT);
      break;
    case DCT_FLIPADST:
      copy_fliplr(src_diff, diff_stride, 16, src_diff2, 16);
      vp10_fht16x16(src_diff2, coeff, 16, DCT_ADST);
      break;
    case FLIPADST_FLIPADST:
      copy_fliplrud(src_diff, diff_stride, 16, src_diff2, 16);
      vp10_fht16x16(src_diff2, coeff, 16, ADST_ADST);
      break;
    case ADST_FLIPADST:
      copy_fliplr(src_diff, diff_stride, 16, src_diff2, 16);
      vp10_fht16x16(src_diff2, coeff, 16, ADST_ADST);
      break;
    case FLIPADST_ADST:
      copy_flipud(src_diff, diff_stride, 16, src_diff2, 16);
      vp10_fht16x16(src_diff2, coeff, 16, ADST_ADST);
      break;
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
    case DST_DST:
    case DCT_DST:
    case DST_DCT:
    case DST_ADST:
    case ADST_DST:
      // Use C version since DST exists only in C
      vp10_fht16x16_c(src_diff, coeff, diff_stride, tx_type);
      break;
    case DST_FLIPADST:
      copy_fliplr(src_diff, diff_stride, 16, src_diff2, 16);
      vp10_fht16x16_c(src_diff2, coeff, 16, DST_ADST);
      break;
    case FLIPADST_DST:
      copy_flipud(src_diff, diff_stride, 16, src_diff2, 16);
      vp10_fht16x16_c(src_diff2, coeff, 16, ADST_DST);
      break;
636
637
638
    case IDTX:
      fwd_idtx_c(src_diff, coeff, diff_stride, 16);
      break;
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
#endif  // CONFIG_EXT_TX
    default:
      assert(0);
      break;
  }
}

static void fwd_txfm_16x16_1(const int16_t *src_diff, tran_low_t *coeff,
                             int diff_stride, TX_TYPE tx_type) {
#if CONFIG_EXT_TX
  int16_t src_diff2[256];
#endif  // CONFIG_EXT_TX
  switch (tx_type) {
    case DCT_DCT:
    case ADST_DCT:
    case DCT_ADST:
    case ADST_ADST:
      vpx_fdct16x16_1(src_diff, coeff, diff_stride);
      break;
#if CONFIG_EXT_TX
    case FLIPADST_DCT:
      copy_flipud(src_diff, diff_stride, 16, src_diff2, 16);
      vp10_fht16x16(src_diff2, coeff, 16, ADST_DCT);
      break;
    case DCT_FLIPADST:
      copy_fliplr(src_diff, diff_stride, 16, src_diff2, 16);
      vp10_fht16x16(src_diff2, coeff, 16, DCT_ADST);
      break;
    case FLIPADST_FLIPADST:
      copy_fliplrud(src_diff, diff_stride, 16, src_diff2, 16);
      vp10_fht16x16(src_diff2, coeff, 16, ADST_ADST);
      break;
    case ADST_FLIPADST:
      copy_fliplr(src_diff, diff_stride, 16, src_diff2, 16);
      vp10_fht16x16(src_diff2, coeff, 16, ADST_ADST);
      break;
    case FLIPADST_ADST:
      copy_flipud(src_diff, diff_stride, 16, src_diff2, 16);
      vp10_fht16x16(src_diff2, coeff, 16, ADST_ADST);
      break;
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
    case DST_DST:
    case DCT_DST:
    case DST_DCT:
    case DST_ADST:
    case ADST_DST:
      // Use C version since DST exists only in C
      vp10_fht16x16_c(src_diff, coeff, diff_stride, tx_type);
      break;
    case DST_FLIPADST:
      copy_fliplr(src_diff, diff_stride, 16, src_diff2, 16);
      vp10_fht16x16_c(src_diff2, coeff, 16, DST_ADST);
      break;
    case FLIPADST_DST:
      copy_flipud(src_diff, diff_stride, 16, src_diff2, 16);
      vp10_fht16x16_c(src_diff2, coeff, 16, ADST_DST);
      break;
695
696
697
    case IDTX:
      fwd_idtx_c(src_diff, coeff, diff_stride, 16);
      break;
698
#endif  // CONFIG_EXT_TX
hui su's avatar
hui su committed
699
700
701
702
703
704
705
706
707
708
709
710
711
    default:
      assert(0);
      break;
  }
}

static void fwd_txfm_32x32(int rd_transform, const int16_t *src_diff,
                           tran_low_t *coeff, int diff_stride,
                           TX_TYPE tx_type) {
  switch (tx_type) {
    case DCT_DCT:
      fdct32x32(rd_transform, src_diff, coeff, diff_stride);
      break;
712
713
714
715
716
#if CONFIG_EXT_TX
    case IDTX:
      fwd_idtx_c(src_diff, coeff, diff_stride, 32);
      break;
#endif  // CONFIG_EXT_TX
hui su's avatar
hui su committed
717
718
719
720
721
722
723
724
725
726
727
    case ADST_DCT:
    case DCT_ADST:
    case ADST_ADST:
      assert(0);
      break;
    default:
      assert(0);
      break;
  }
}

728
729
730
731
732
733
734
static void fwd_txfm_32x32_1(const int16_t *src_diff,
                             tran_low_t *coeff, int diff_stride,
                             TX_TYPE tx_type) {
  switch (tx_type) {
    case DCT_DCT:
      vpx_fdct32x32_1(src_diff, coeff, diff_stride);
      break;
735
736
737
738
739
#if CONFIG_EXT_TX
    case IDTX:
      fwd_idtx_c(src_diff, coeff, diff_stride, 32);
      break;
#endif  // CONFIG_EXT_TX
740
741
742
743
744
745
746
747
748
749
750
    case ADST_DCT:
    case DCT_ADST:
    case ADST_ADST:
      assert(0);
      break;
    default:
      assert(0);
      break;
  }
}

hui su's avatar
hui su committed
751
752
#if CONFIG_VP9_HIGHBITDEPTH
void vp10_highbd_fwd_txfm_4x4(const int16_t *src_diff, tran_low_t *coeff,
753
754
755
756
757
                              int diff_stride, TX_TYPE tx_type, int lossless) {
  if (lossless) {
    assert(tx_type == DCT_DCT);
    vp10_highbd_fwht4x4(src_diff, coeff, diff_stride);
  } else {
758
759
760
#if CONFIG_EXT_TX
  int16_t src_diff2[16];
#endif  // CONFIG_EXT_TX
761
762
    switch (tx_type) {
      case DCT_DCT:
763
      vpx_highbd_fdct4x4(src_diff, coeff, diff_stride);
764
765
766
767
768
769
        break;
      case ADST_DCT:
      case DCT_ADST:
      case ADST_ADST:
        vp10_highbd_fht4x4(src_diff, coeff, diff_stride, tx_type);
        break;
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
#if CONFIG_EXT_TX
    case FLIPADST_DCT:
      copy_flipud(src_diff, diff_stride, 4, src_diff2, 4);
      vp10_highbd_fht4x4(src_diff2, coeff, 4, ADST_DCT);
      break;
    case DCT_FLIPADST:
      copy_fliplr(src_diff, diff_stride, 4, src_diff2, 4);
      vp10_highbd_fht4x4(src_diff2, coeff, 4, DCT_ADST);
      break;
    case FLIPADST_FLIPADST:
      copy_fliplrud(src_diff, diff_stride, 4, src_diff2, 4);
      vp10_highbd_fht4x4(src_diff2, coeff, 4, ADST_ADST);
      break;
    case ADST_FLIPADST:
      copy_fliplr(src_diff, diff_stride, 4, src_diff2, 4);
      vp10_highbd_fht4x4(src_diff2, coeff, 4, ADST_ADST);
      break;
    case FLIPADST_ADST:
      copy_flipud(src_diff, diff_stride, 4, src_diff2, 4);
      vp10_highbd_fht4x4(src_diff2, coeff, 4, ADST_ADST);
      break;
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
    case DST_DST:
    case DCT_DST:
    case DST_DCT:
    case DST_ADST:
    case ADST_DST:
      // Use C version since DST exists only in C
      vp10_highbd_fht4x4_c(src_diff, coeff, diff_stride, tx_type);
      break;
    case DST_FLIPADST:
      copy_fliplr(src_diff, diff_stride, 4, src_diff2, 4);
      vp10_highbd_fht4x4_c(src_diff2, coeff, 4, DST_ADST);
      break;
    case FLIPADST_DST:
      copy_flipud(src_diff, diff_stride, 4, src_diff2, 4);
      vp10_highbd_fht4x4_c(src_diff2, coeff, 4, ADST_DST);
      break;
807
808
809
    case IDTX:
      fwd_idtx_c(src_diff, coeff, diff_stride, 4);
      break;
810
#endif  // CONFIG_EXT_TX
811
812
813
      default:
        assert(0);
        break;
814
    }
hui su's avatar
hui su committed
815
816
817
818
  }
}

static void highbd_fwd_txfm_8x8(const int16_t *src_diff, tran_low_t *coeff,
819
820
821
822
                                int diff_stride, TX_TYPE tx_type) {
#if CONFIG_EXT_TX
  int16_t src_diff2[64];
#endif  // CONFIG_EXT_TX
hui su's avatar
hui su committed
823
824
825
826
827
828
829
830
831
  switch (tx_type) {
    case DCT_DCT:
      vpx_highbd_fdct8x8(src_diff, coeff, diff_stride);
      break;
    case ADST_DCT:
    case DCT_ADST:
    case ADST_ADST:
      vp10_highbd_fht8x8(src_diff, coeff, diff_stride, tx_type);
      break;
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
#if CONFIG_EXT_TX
    case FLIPADST_DCT:
      copy_flipud(src_diff, diff_stride, 8, src_diff2, 8);
      vp10_highbd_fht8x8(src_diff2, coeff, 8, ADST_DCT);
      break;
    case DCT_FLIPADST:
      copy_fliplr(src_diff, diff_stride, 8, src_diff2, 8);
      vp10_highbd_fht8x8(src_diff2, coeff, 8, DCT_ADST);
      break;
    case FLIPADST_FLIPADST:
      copy_fliplrud(src_diff, diff_stride, 8, src_diff2, 8);
      vp10_highbd_fht8x8(src_diff2, coeff, 8, ADST_ADST);
      break;
    case ADST_FLIPADST:
      copy_fliplr(src_diff, diff_stride, 8, src_diff2, 8);
      vp10_highbd_fht8x8(src_diff2, coeff, 8, ADST_ADST);
      break;
    case FLIPADST_ADST:
      copy_flipud(src_diff, diff_stride, 8, src_diff2, 8);
      vp10_highbd_fht8x8(src_diff2, coeff, 8, ADST_ADST);
      break;
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
    case DST_DST:
    case DCT_DST:
    case DST_DCT:
    case DST_ADST:
    case ADST_DST:
      // Use C version since DST exists only in C
      vp10_highbd_fht8x8_c(src_diff, coeff, diff_stride, tx_type);
      break;
    case DST_FLIPADST:
      copy_fliplr(src_diff, diff_stride, 8, src_diff2, 8);
      vp10_highbd_fht8x8_c(src_diff2, coeff, 8, DST_ADST);
      break;
    case FLIPADST_DST:
      copy_flipud(src_diff, diff_stride, 8, src_diff2, 8);
      vp10_highbd_fht8x8_c(src_diff2, coeff, 8, ADST_DST);
      break;
869
870
871
    case IDTX:
      fwd_idtx_c(src_diff, coeff, diff_stride, 8);
      break;
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
#endif  // CONFIG_EXT_TX
    default:
      assert(0);
      break;
  }
}

static void highbd_fwd_txfm_8x8_1(const int16_t *src_diff, tran_low_t *coeff,
                                  int diff_stride, TX_TYPE tx_type) {
#if CONFIG_EXT_TX
  int16_t src_diff2[64];
#endif  // CONFIG_EXT_TX
  switch (tx_type) {
    case DCT_DCT:
      vpx_highbd_fdct8x8_1(src_diff, coeff, diff_stride);
      break;
    case ADST_DCT:
    case DCT_ADST:
    case ADST_ADST:
      vp10_highbd_fht8x8(src_diff, coeff, diff_stride, tx_type);
      break;
#if CONFIG_EXT_TX
    case FLIPADST_DCT:
      copy_flipud(src_diff, diff_stride, 8, src_diff2, 8);
      vp10_highbd_fht8x8(src_diff2, coeff, 8, ADST_DCT);
      break;
    case DCT_FLIPADST:
      copy_fliplr(src_diff, diff_stride, 8, src_diff2, 8);
      vp10_highbd_fht8x8(src_diff2, coeff, 8, DCT_ADST);
      break;
    case FLIPADST_FLIPADST:
      copy_fliplrud(src_diff, diff_stride, 8, src_diff2, 8);
      vp10_highbd_fht8x8(src_diff2, coeff, 8, ADST_ADST);
      break;
    case ADST_FLIPADST:
      copy_fliplr(src_diff, diff_stride, 8, src_diff2, 8);
      vp10_highbd_fht8x8(src_diff2, coeff, 8, ADST_ADST);
      break;
    case FLIPADST_ADST:
      copy_flipud(src_diff, diff_stride, 8, src_diff2, 8);
      vp10_highbd_fht8x8(src_diff2, coeff, 8, ADST_ADST);
      break;
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
    case DST_DST:
    case DCT_DST:
    case DST_DCT:
    case DST_ADST:
    case ADST_DST:
      // Use C version since DST exists only in C
      vp10_highbd_fht8x8_c(src_diff, coeff, diff_stride, tx_type);
      break;
    case DST_FLIPADST:
      copy_fliplr(src_diff, diff_stride, 8, src_diff2, 8);
      vp10_highbd_fht8x8_c(src_diff2, coeff, 8, DST_ADST);
      break;
    case FLIPADST_DST:
      copy_flipud(src_diff, diff_stride, 8, src_diff2, 8);
      vp10_highbd_fht8x8_c(src_diff2, coeff, 8, ADST_DST);
      break;
930
931
932
    case IDTX:
      fwd_idtx_c(src_diff, coeff, diff_stride, 8);
      break;
933
#endif  // CONFIG_EXT_TX
hui su's avatar
hui su committed
934
935
936
937
938
939
940
    default:
      assert(0);
      break;
  }
}

static void highbd_fwd_txfm_16x16(const int16_t *src_diff, tran_low_t *coeff,
941
942
943
944
                                  int diff_stride, TX_TYPE tx_type) {
#if CONFIG_EXT_TX
  int16_t src_diff2[256];
#endif  // CONFIG_EXT_TX
hui su's avatar
hui su committed
945
946
947
948
949
950
951
952
953
  switch (tx_type) {
    case DCT_DCT:
      vpx_highbd_fdct16x16(src_diff, coeff, diff_stride);
      break;
    case ADST_DCT:
    case DCT_ADST:
    case ADST_ADST:
      vp10_highbd_fht16x16(src_diff, coeff, diff_stride, tx_type);
      break;
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
#if CONFIG_EXT_TX
    case FLIPADST_DCT:
      copy_flipud(src_diff, diff_stride, 16, src_diff2, 16);
      vp10_highbd_fht16x16(src_diff2, coeff, 16, ADST_DCT);
      break;
    case DCT_FLIPADST:
      copy_fliplr(src_diff, diff_stride, 16, src_diff2, 16);
      vp10_highbd_fht16x16(src_diff2, coeff, 16, DCT_ADST);
      break;
    case FLIPADST_FLIPADST:
      copy_fliplrud(src_diff, diff_stride, 16, src_diff2, 16);
      vp10_highbd_fht16x16(src_diff2, coeff, 16, ADST_ADST);
      break;
    case ADST_FLIPADST:
      copy_fliplr(src_diff, diff_stride, 16, src_diff2, 16);
      vp10_highbd_fht16x16(src_diff2, coeff, 16, ADST_ADST);
      break;
    case FLIPADST_ADST:
      copy_flipud(src_diff, diff_stride, 16, src_diff2, 16);
      vp10_highbd_fht16x16(src_diff2, coeff, 16, ADST_ADST);
      break;
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
    case DST_DST:
    case DCT_DST:
    case DST_DCT:
    case DST_ADST:
    case ADST_DST:
      // Use C version since DST exists only in C
      vp10_highbd_fht16x16_c(src_diff, coeff, diff_stride, tx_type);
      break;
    case DST_FLIPADST:
      copy_fliplr(src_diff, diff_stride, 16, src_diff2, 16);
      vp10_highbd_fht16x16_c(src_diff2, coeff, 16, DST_ADST);
      break;
    case FLIPADST_DST:
      copy_flipud(src_diff, diff_stride, 16, src_diff2, 16);
      vp10_highbd_fht16x16_c(src_diff2, coeff, 16, ADST_DST);
      break;
991
992
993
    case IDTX:
      fwd_idtx_c(src_diff, coeff, diff_stride, 16);
      break;
994
995
996
997
998
999
1000
#endif  // CONFIG_EXT_TX
    default:
      assert(0);
      break;
  }
}

For faster browsing, not all history is shown. View entire blame