encodemb.c 63.3 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
568
569
570
571
572
573
574
575
576
    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;
577
578
579
    case IDTX:
      fwd_idtx_c(src_diff, coeff, diff_stride, 8);
      break;
580
#endif  // CONFIG_EXT_TX
hui su's avatar
hui su committed
581
582
583
584
585
586
587
588
    default:
      assert(0);
      break;
  }
}

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
712
713
714
715
716
717
718
719
720
721
722
    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;
    case ADST_DCT:
    case DCT_ADST:
    case ADST_ADST:
      assert(0);
      break;
    default:
      assert(0);
      break;
  }
}

723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
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;
    case ADST_DCT:
    case DCT_ADST:
    case ADST_ADST:
      assert(0);
      break;
    default:
      assert(0);
      break;
  }
}

hui su's avatar
hui su committed
741
742
#if CONFIG_VP9_HIGHBITDEPTH
void vp10_highbd_fwd_txfm_4x4(const int16_t *src_diff, tran_low_t *coeff,
743
744
745
746
747
                              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 {
748
749
750
#if CONFIG_EXT_TX
  int16_t src_diff2[16];
#endif  // CONFIG_EXT_TX
hui su's avatar
hui su committed
751
752
  switch (tx_type) {
    case DCT_DCT:
753
      vpx_highbd_fdct4x4(src_diff, coeff, diff_stride);
hui su's avatar
hui su committed
754
755
756
757
758
759
      break;
    case ADST_DCT:
    case DCT_ADST:
    case ADST_ADST:
      vp10_highbd_fht4x4(src_diff, coeff, diff_stride, tx_type);
      break;
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
#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;
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
    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;
797
798
799
    case IDTX:
      fwd_idtx_c(src_diff, coeff, diff_stride, 4);
      break;
800
#endif  // CONFIG_EXT_TX
hui su's avatar
hui su committed
801
802
803
    default:
      assert(0);
      break;
804
    }
hui su's avatar
hui su committed
805
806
807
808
  }
}

static void highbd_fwd_txfm_8x8(const int16_t *src_diff, tran_low_t *coeff,
809
810
811
812
                                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
813
814
815
816
817
818
819
820
821
  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;
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
#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;
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
    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;
859
860
861
    case IDTX:
      fwd_idtx_c(src_diff, coeff, diff_stride, 8);
      break;
862
863
864
865
866
867
868
869
870
871
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
#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;
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
    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;
920
921
922
    case IDTX:
      fwd_idtx_c(src_diff, coeff, diff_stride, 8);
      break;
923
#endif  // CONFIG_EXT_TX
hui su's avatar
hui su committed
924
925
926
927
928
929
930
    default:
      assert(0);
      break;
  }
}

static void highbd_fwd_txfm_16x16(const int16_t *src_diff, tran_low_t *coeff,
931
932
933
934
                                  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
935
936
937
938
939
940
941
942
943
  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;
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
#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;
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
    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;
981
982
983
    case IDTX:
      fwd_idtx_c(src_diff, coeff, diff_stride, 16);
      break;
984
985
986
987
988
989
990
991
992
993
994
995
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
1025
#endif  // CONFIG_EXT_TX
    default:
      assert(0);
      break;
  }
}

static void highbd_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:
      vpx_highbd_fdct16x16_1(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;
#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;
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
    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;
1042
1043
1044
    case IDTX:
      fwd_idtx_c(src_diff, coeff, diff_stride, 16);
      break;
1045
#endif  // CONFIG_EXT_TX
hui su's avatar
hui su committed
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
    default:
      assert(0);
      break;
  }
}

static void highbd_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:
      highbd_fdct32x32(rd_transform, src_diff, coeff, diff_stride);
      break;
    case ADST_DCT:
    case DCT_ADST:
    case ADST_ADST:
      assert(0);
      break;
    default:
      assert(0);
      break;
  }
}
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086

static void highbd_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_highbd_fdct32x32_1(src_diff, coeff, diff_stride);
      break;
    case ADST_DCT:
    case DCT_ADST:
    case ADST_ADST:
      assert(0);
      break;
    default:
      assert(0);
      break;
  }
}
hui su's avatar
hui su committed
1087
1088
#endif  // CONFIG_VP9_HIGHBITDEPTH

1089
void vp10_xform_quant_fp(MACROBLOCK *x, int plane, int block,
1090
1091
                         int blk_row, int blk_col,
                         BLOCK_SIZE plane_bsize, TX_SIZE tx_size) {
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
  MACROBLOCKD *const xd = &x->e_mbd;
  const struct macroblock_plane *const p = &x->plane[plane];
  const struct macroblockd_plane *const pd = &xd->plane[plane];
  PLANE_TYPE plane_type = (plane == 0) ? PLANE_TYPE_Y : PLANE_TYPE_UV;
  TX_TYPE tx_type = get_tx_type(plane_type, xd, block, tx_size);
  const scan_order *const scan_order =
      get_scan(tx_size, tx_type, is_inter_block(&xd->mi[0]->mbmi));
  tran_low_t *const coeff = BLOCK_OFFSET(p->coeff, block);
  tran_low_t *const qcoeff = BLOCK_OFFSET(p->qcoeff, block);
  tran_low_t *const dqcoeff = BLOCK_OFFSET(pd->dqcoeff, block);
  uint16_t *const eob = &p->eobs[block];
  const int diff_stride = 4 * num_4x4_blocks_wide_lookup[plane_bsize];
  const int16_t *src_diff;
1105
  src_diff = &p->src_diff[4 * (blk_row * diff_stride + blk_col)];
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134

#if CONFIG_VP9_HIGHBITDEPTH
  if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
    switch (tx_size) {
      case TX_32X32:
        highbd_fwd_txfm_32x32(x->use_lp32x32fdct, src_diff, coeff, diff_stride,
                              tx_type);
        vp10_highbd_quantize_fp_32x32(coeff, 1024, x->skip_block, p->zbin,
                                      p->round_fp, p->quant_fp, p->quant_shift,
                                      qcoeff, dqcoeff, pd->dequant,
                                      eob, scan_order->scan,
                                      scan_order->iscan);
        break;
      case TX_16X16:
        highbd_fwd_txfm_16x16(src_diff, coeff, diff_stride, tx_type);
        vp10_highbd_quantize_fp(coeff, 256, x->skip_block, p->zbin, p->round_fp,
                                p->quant_fp, p->quant_shift, qcoeff, dqcoeff,
                                pd->dequant, eob,
                                scan_order->scan, scan_order->iscan);
        break;
      case TX_8X8:
        highbd_fwd_txfm_8x8(src_diff, coeff, diff_stride, tx_type);
        vp10_highbd_quantize_fp(coeff, 64, x->skip_block, p->zbin, p->round_fp,
                                p->quant_fp, p->quant_shift, qcoeff, dqcoeff,
                                pd->dequant, eob,
                                scan_order->scan, scan_order->iscan);
        break;
      case TX_4X4:
        vp10_highbd_fwd_txfm_4x4(src_diff, coeff, diff_stride, tx_type,
1135
                                 xd->lossless);
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
        vp10_highbd_quantize_fp(coeff, 16, x->skip_block, p->zbin, p->round_fp,
                                p->quant_fp, p->quant_shift, qcoeff, dqcoeff,
                                pd->dequant, eob,
                                scan_order->scan, scan_order->iscan);
        break;
      default:
        assert(0);
    }
    return;
  }
#endif  // CONFIG_VP9_HIGHBITDEPTH

  switch (tx_size) {
    case TX_32X32:
      fwd_txfm_32x32(x->use_lp32x32fdct, src_diff, coeff, diff_stride, tx_type);
      vp10_quantize_fp_32x32(coeff, 1024, x->skip_block, p->zbin, p->round_fp,
                             p->quant_fp, p->quant_shift, qcoeff, dqcoeff,
                             pd->dequant, eob, scan_order->scan,
                             scan_order->iscan);
      break;
    case TX_16X16:
      fwd_txfm_16x16(src_diff, coeff, diff_stride, tx_type);
      vp10_quantize_fp(coeff, 256, x->skip_block, p->zbin, p->round_fp,
                       p->quant_fp, p->quant_shift, qcoeff, dqcoeff,
                       pd->dequant, eob,
                       scan_order->scan, scan_order->iscan);
      break;
    case TX_8X8:
      fwd_txfm_8x8(src_diff, coeff, diff_stride, tx_type);
      vp10_quantize_fp(coeff, 64, x->skip_block, p->zbin, p->round_fp,
                       p->quant_fp, p->quant_shift, qcoeff, dqcoeff,
                       pd->dequant, eob,
                       scan_order->scan, scan_order->iscan);
      break;
    case TX_4X4:
      vp10_fwd_txfm_4x4(src_diff, coeff, diff_stride, tx_type,
1172
                        xd->lossless);
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
      vp10_quantize_fp(coeff, 16, x->skip_block, p->zbin, p->round_fp,
                       p->quant_fp, p->quant_shift, qcoeff, dqcoeff,
                       pd->dequant, eob,
                       scan_order->scan, scan_order->iscan);
      break;
    default:
      assert(0);
      break;
  }
}

void vp10_xform_quant_dc(MACROBLOCK *x, int plane, int block,
1185
1186
                         int blk_row, int blk_col,
                         BLOCK_SIZE plane_bsize, TX_SIZE tx_size) {
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
  MACROBLOCKD *const xd = &x->e_mbd;
  const struct macroblock_plane *const p = &x->plane[plane];
  const struct macroblockd_plane *const pd = &xd->plane[plane];
  PLANE_TYPE plane_type = (plane == 0) ? PLANE_TYPE_Y : PLANE_TYPE_UV;
  TX_TYPE tx_type = get_tx_type(plane_type, xd, block, tx_size);
  tran_low_t *const coeff = BLOCK_OFFSET(p->coeff, block);
  tran_low_t *const qcoeff = BLOCK_OFFSET(p->qcoeff, block);
  tran_low_t *const dqcoeff = BLOCK_OFFSET(pd->dqcoeff, block);
  uint16_t *const eob = &p->eobs[block];
  const int diff_stride = 4 * num_4x4_blocks_wide_lookup[plane_bsize];
  const int16_t *src_diff;
1198
  src_diff = &p->src_diff[4 * (blk_row * diff_stride + blk_col)];
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222

#if CONFIG_VP9_HIGHBITDEPTH
  if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
    switch (tx_size) {
      case TX_32X32:
        highbd_fwd_txfm_32x32_1(src_diff, coeff, diff_stride, tx_type);
        vpx_highbd_quantize_dc_32x32(coeff, x->skip_block, p->round,
                                     p->quant_fp[0], qcoeff, dqcoeff,
                                     pd->dequant[0], eob);
        break;
      case TX_16X16:
        highbd_fwd_txfm_16x16_1(src_diff, coeff, diff_stride, tx_type);
        vpx_highbd_quantize_dc(coeff, 256, x->skip_block, p->round,
                               p->quant_fp[0], qcoeff, dqcoeff,
                               pd->dequant[0], eob);
        break;
      case TX_8X8:
        highbd_fwd_txfm_8x8_1(src_diff, coeff, diff_stride, tx_type);
        vpx_highbd_quantize_dc(coeff, 64, x->skip_block, p->round,
                               p->quant_fp[0], qcoeff, dqcoeff,
                               pd->dequant[0], eob);
        break;
      case TX_4X4:
        vp10_highbd_fwd_txfm_4x4(src_diff, coeff, diff_stride, tx_type,
1223
                                 xd->lossless);
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
        vpx_highbd_quantize_dc(coeff, 16, x->skip_block, p->round,
                               p->quant_fp[0], qcoeff, dqcoeff,
                               pd->dequant[0], eob);
        break;
      default:
        assert(0);
    }
    return;
  }
#endif  // CONFIG_VP9_HIGHBITDEPTH

  switch (tx_size) {
    case TX_32X32:
      fwd_txfm_32x32_1(src_diff, coeff, diff_stride, tx_type);
      vpx_quantize_dc_32x32(coeff, x->skip_block, p->round,
                            p->quant_fp[0], qcoeff, dqcoeff,
                            pd->dequant[0], eob);
      break;
    case TX_16X16:
      fwd_txfm_16x16_1(src_diff, coeff, diff_stride, tx_type);
      vpx_quantize_dc(coeff, 256, x->skip_block, p->round,
                     p->quant_fp[0], qcoeff, dqcoeff,
                     pd->dequant[0], eob);
      break;
    case TX_8X8:
      fwd_txfm_8x8_1(src_diff, coeff, diff_stride, tx_type);
      vpx_quantize_dc(coeff, 64, x->skip_block, p->round,
                      p->quant_fp[0], qcoeff, dqcoeff,
                      pd->dequant[0], eob);
      break;
    case TX_4X4:
      vp10_fwd_txfm_4x4(src_diff, coeff, diff_stride, tx_type,
1256
                        xd->lossless);
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
      vpx_quantize_dc(coeff, 16, x->skip_block, p->round,
                      p->quant_fp[0], qcoeff, dqcoeff,
                      pd->dequant[0], eob);
      break;
    default:
      assert(0);
      break;
  }
}

Jingning Han's avatar
Jingning Han committed
1267
void vp10_xform_quant(MACROBLOCK *x, int plane, int block,
1268
1269
                      int blk_row, int blk_col,
                      BLOCK_SIZE plane_bsize, TX_SIZE tx_size) {
Jingning Han's avatar
Jingning Han committed
1270
1271
1272
  MACROBLOCKD *const xd = &x->e_mbd;
  const struct macroblock_plane *const p = &x->plane[plane];
  const struct macroblockd_plane *const pd = &xd->plane[plane];
hui su's avatar
hui su committed
1273
  PLANE_TYPE plane_type = (plane == 0) ? PLANE_TYPE_Y : PLANE_TYPE_UV;
hui su's avatar
hui su committed
1274
  TX_TYPE tx_type = get_tx_type(plane_type, xd, block, tx_size);
1275
1276
  const scan_order *const scan_order =
      get_scan(tx_size, tx_type, is_inter_block(&xd->mi[0]->mbmi));
Jingning Han's avatar
Jingning Han committed
1277
1278
1279
1280
1281
1282
  tran_low_t *const coeff = BLOCK_OFFSET(p->coeff, block);
  tran_low_t *const qcoeff = BLOCK_OFFSET(p->qcoeff, block);
  tran_low_t *const dqcoeff = BLOCK_OFFSET(pd->dqcoeff, block);
  uint16_t *const eob = &p->eobs[block];
  const int diff_stride = 4 * num_4x4_blocks_wide_lookup[plane_bsize];
  const int16_t *src_diff;
1283
  src_diff = &p->src_diff[4 * (blk_row * diff_stride + blk_col)];
Jingning Han's avatar
Jingning Han committed
1284
1285
1286
1287
1288

#if CONFIG_VP9_HIGHBITDEPTH
  if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
     switch (tx_size) {
      case TX_32X32:
hui su's avatar
hui su committed
1289
        highbd_fwd_txfm_32x32(x->use_lp32x32fdct, src_diff, coeff, diff_stride,
1290
                              tx_type);
Jingning Han's avatar
Jingning Han committed
1291
1292
1293
1294
1295
1296
        vpx_highbd_quantize_b_32x32(coeff, 1024, x->skip_block, p->zbin,
                                    p->round, p->quant, p->quant_shift, qcoeff,
                                    dqcoeff, pd->dequant, eob,
                                    scan_order->scan, scan_order->iscan);
        break;
      case TX_16X16:
hui su's avatar
hui su committed
1297
        highbd_fwd_txfm_16x16(src_diff, coeff, diff_stride, tx_type);
Jingning Han's avatar
Jingning Han committed
1298
1299
1300
1301
1302
1303
        vpx_highbd_quantize_b(coeff, 256, x->skip_block, p->zbin, p->round,
                              p->quant, p->quant_shift, qcoeff, dqcoeff,
                              pd->dequant, eob,
                              scan_order->scan, scan_order->iscan);
        break;
      case TX_8X8:
hui su's avatar
hui su committed
1304
        highbd_fwd_txfm_8x8(src_diff, coeff, diff_stride, tx_type);
Jingning Han's avatar
Jingning Han committed
1305
1306
1307
1308
1309
1310
        vpx_highbd_quantize_b(coeff, 64, x->skip_block, p->zbin, p->round,
                              p->quant, p->quant_shift, qcoeff, dqcoeff,
                              pd->dequant, eob,
                              scan_order->scan, scan_order->iscan);
        break;
      case TX_4X4:
hui su's avatar
hui su committed
1311
        vp10_highbd_fwd_txfm_4x4(src_diff, coeff, diff_stride, tx_type,
1312
                                 xd->lossless);
Jingning Han's avatar
Jingning Han committed
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
        vpx_highbd_quantize_b(coeff, 16, x->skip_block, p->zbin, p->round,
                              p->quant, p->quant_shift, qcoeff, dqcoeff,
                              pd->dequant, eob,
                              scan_order->scan, scan_order->iscan);
        break;
      default:
        assert(0);
    }
    return;
  }
#endif  // CONFIG_VP9_HIGHBITDEPTH

  switch (tx_size) {
    case TX_32X32:
hui su's avatar
hui su committed
1327
      fwd_txfm_32x32(x->use_lp32x32fdct, src_diff, coeff, diff_stride, tx_type);
Jingning Han's avatar
Jingning Han committed
1328
1329
1330
1331
1332
1333
      vpx_quantize_b_32x32(coeff, 1024, x->skip_block, p->zbin, p->round,
                           p->quant, p->quant_shift, qcoeff, dqcoeff,
                           pd->dequant, eob, scan_order->scan,
                           scan_order->iscan);
      break;
    case TX_16X16:
hui su's avatar
hui su committed
1334
      fwd_txfm_16x16(src_diff, coeff, diff_stride, tx_type);
Jingning Han's avatar
Jingning Han committed
1335
1336
1337
1338
1339
1340
      vpx_quantize_b(coeff, 256, x->skip_block, p->zbin, p->round,
                     p->quant, p->quant_shift, qcoeff, dqcoeff,
                     pd->dequant, eob,
                     scan_order->scan, scan_order->iscan);
      break;
    case TX_8X8:
hui su's avatar
hui su committed
1341
      fwd_txfm_8x8(src_diff, coeff, diff_stride, tx_type);
Jingning Han's avatar
Jingning Han committed
1342
1343
1344
1345
1346
1347
      vpx_quantize_b(coeff, 64, x->skip_block, p->zbin, p->round,
                     p->quant, p->quant_shift, qcoeff, dqcoeff,
                     pd->dequant, eob,
                     scan_order->scan, scan_order->iscan);
      break;
    case TX_4X4:
1348
      vp10_fwd_txfm_4x4(src_diff, coeff, diff_stride, tx_type, xd->lossless);
Jingning Han's avatar
Jingning Han committed
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
      vpx_quantize_b(coeff, 16, x->skip_block, p->zbin, p->round,
                     p->quant, p->quant_shift, qcoeff, dqcoeff,
                     pd->dequant, eob,
                     scan_order->scan, scan_order->iscan);
      break;
    default:
      assert(0);
      break;
  }
}

1360
1361
static void encode_block(int plane, int block, int blk_row, int blk_col,
                         BLOCK_SIZE plane_bsize,
Jingning Han's avatar
Jingning Han committed
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
                         TX_SIZE tx_size, void *arg) {
  struct encode_b_args *const args = arg;
  MACROBLOCK *const x = args->x;
  MACROBLOCKD *const xd = &x->e_mbd;
  struct optimize_ctx *const ctx = args->ctx;
  struct macroblock_plane *const p = &x->plane[plane];
  struct macroblockd_plane *const pd = &xd->plane[plane];
  tran_low_t *const dqcoeff = BLOCK_OFFSET(pd->dqcoeff, block);
  uint8_t *dst;
  ENTROPY_CONTEXT *a, *l;
hui su's avatar
hui su committed
1372
  TX_TYPE tx_type = get_tx_type(pd->plane_type, xd, block, tx_size);
1373
1374
1375
#if CONFIG_VAR_TX
  int i;
#endif
1376
1377
1378
  dst = &pd->dst.buf[4 * blk_row * pd->dst.stride + 4 * blk_col];
  a = &ctx->ta[plane][blk_col];
  l = &ctx->tl[plane][blk_row];
Jingning Han's avatar
Jingning Han committed
1379
1380
1381

  // TODO(jingning): per transformed block zero forcing only enabled for
  // luma component. will integrate chroma components as well.
1382
1383
1384
1385
1386
1387
1388
  // Turn this back on when the rate-distortion loop is synchronized with
  // the recursive transform block coding.
//  if (x->zcoeff_blk[tx_size][block] && plane == 0) {
//    p->eobs[block] = 0;
//    *a = *l = 0;
//    return;
//  }
Jingning Han's avatar
Jingning Han committed
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398

  if (!x->skip_recode) {
    if (x->quant_fp) {
      // Encoding process for rtc mode
      if (x->skip_txfm[0] == SKIP_TXFM_AC_DC && plane == 0) {
        // skip forward transform
        p->eobs[block] = 0;
        *a = *l = 0;
        return;
      } else {
1399
1400
        vp10_xform_quant_fp(x, plane, block, blk_row, blk_col,
                            plane_bsize, tx_size);
Jingning Han's avatar
Jingning Han committed
1401
1402
1403
1404
1405
1406
      }
    } else {
      if (max_txsize_lookup[plane_bsize] == tx_size) {
        int txfm_blk_index = (plane << 2) + (block >> (tx_size << 1));
        if (x->skip_txfm[txfm_blk_index] == SKIP_TXFM_NONE) {
          // full forward transform and quantization
1407
1408
          vp10_xform_quant(x, plane, block, blk_row, blk_col,
                           plane_bsize, tx_size);
Jingning Han's avatar
Jingning Han committed
1409
1410
        } else if (x->skip_txfm[txfm_blk_index] == SKIP_TXFM_AC_ONLY) {
          // fast path forward transform and quantization
1411
1412
          vp10_xform_quant_dc(x, plane, block, blk_row, blk_col,
                              plane_bsize, tx_size);
Jingning Han's avatar
Jingning Han committed
1413
1414
1415
1416
1417
1418
1419
        } else {
          // skip forward transform
          p->eobs[block] = 0;
          *a = *l = 0;
          return;
        }
      } else {
1420
1421
        vp10_xform_quant(x, plane, block, blk_row, blk_col,
                         plane_bsize, tx_size);
Jingning Han's avatar
Jingning Han committed
1422
1423
1424
1425
1426
      }
    }
  }

  if (x->optimize && (!x->skip_recode || !x->skip_optimize)) {
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
    int ctx;
#if CONFIG_VAR_TX
    switch (tx_size) {
      case TX_4X4:
        break;
      case TX_8X8:
        a[0] = !!*(const uint16_t *)&a[0];
        l[0] = !!*(const uint16_t *)&l[0];
        break;
      case TX_16X16:
        a[0] = !!*(const uint32_t *)&a[0];
        l[0] = !!*(const uint32_t *)&l[0];
        break;
      case TX_32X32:
        a[0] = !!*(const uint64_t *)&a[0];
        l[0] = !!*(const uint64_t *)&l[0];
        break;
      default:
        assert(0 && "Invalid transform size.");
        break;
    }
#endif
    ctx = combine_entropy_contexts(*a, *l);
Jingning Han's avatar
Jingning Han committed
1450
1451
1452
1453
1454
    *a = *l = optimize_b(x, plane, block, tx_size, ctx) > 0;
  } else {
    *a = *l = p->eobs[block] > 0;
  }

1455
1456
1457
1458
1459
1460
1461
#if CONFIG_VAR_TX
  for (i = 0; i < (1 << tx_size); ++i) {
    a[i] = a[0];
    l[i] = l[0];
  }
#endif

Jingning Han's avatar
Jingning Han committed
1462
1463
1464
  if (p->eobs[block])
    *(args->skip) = 0;

1465
  if (p->eobs[block] == 0)
Jingning Han's avatar
Jingning Han committed
1466
1467
1468
1469
1470
    return;
#if CONFIG_VP9_HIGHBITDEPTH
  if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
    switch (tx_size) {
      case TX_32X32:
hui su's avatar
hui su committed
1471
1472
        vp10_highbd_inv_txfm_add_32x32(dqcoeff, dst, pd->dst.stride,
                                       p->eobs[block], xd->bd, tx_type);
Jingning Han's avatar
Jingning Han committed
1473
1474
        break;
      case TX_16X16:
hui su's avatar
hui su committed
1475
1476
        vp10_highbd_inv_txfm_add_16x16(dqcoeff, dst, pd->dst.stride,
                                       p->eobs[block], xd->bd, tx_type);
Jingning Han's avatar
Jingning Han committed
1477
1478
        break;
      case TX_8X8:
hui su's avatar
hui su committed
1479
1480
        vp10_highbd_inv_txfm_add_8x8(dqcoeff, dst, pd->dst.stride,
                                     p->eobs[block], xd->bd, tx_type);
Jingning Han's avatar
Jingning Han committed
1481
1482
1483
1484
1485
        break;
      case TX_4X4:
        // this is like vp10_short_idct4x4 but has a special case around eob<=1
        // which is significant (not just an optimization) for the lossless
        // case.
hui su's avatar
hui su committed
1486
1487
        vp10_highbd_inv_txfm_add_4x4(dqcoeff, dst, pd->dst.stride,
                                     p->eobs[block], xd->bd, tx_type,
1488
                                     xd->lossless);
Jingning Han's avatar
Jingning Han committed
1489
1490
1491
        break;
      default:
        assert(0 && "Invalid transform size");
hui su's avatar
hui su committed
1492
        break;
Jingning Han's avatar
Jingning Han committed
1493
    }
hui su's avatar
hui su committed
1494

Jingning Han's avatar
Jingning Han committed
1495
1496
1497
1498
1499
1500
    return;
  }
#endif  // CONFIG_VP9_HIGHBITDEPTH

  switch (tx_size) {
    case TX_32X32:
hui su's avatar
hui su committed
1501
1502
      vp10_inv_txfm_add_32x32(dqcoeff, dst, pd->dst.stride, p->eobs[block],
                              tx_type);
Jingning Han's avatar
Jingning Han committed
1503
1504
      break;
    case TX_16X16:
hui su's avatar
hui su committed
1505
1506
      vp10_inv_txfm_add_16x16(dqcoeff, dst, pd->dst.stride, p->eobs[block],
                              tx_type);
Jingning Han's avatar
Jingning Han committed
1507
1508
      break;
    case TX_8X8:
hui su's avatar
hui su committed
1509
1510
      vp10_inv_txfm_add_8x8(dqcoeff, dst, pd->dst.stride, p->eobs[block],
                            tx_type);
Jingning Han's avatar
Jingning Han committed
1511
1512
1513
1514
1515
      break;
    case TX_4X4:
      // this is like vp10_short_idct4x4 but has a special case around eob<=1
      // which is significant (not just an optimization) for the lossless
      // case.
hui su's avatar
hui su committed
1516
      vp10_inv_txfm_add_4x4(dqcoeff, dst, pd->dst.stride, p->eobs[block],
1517
                            tx_type, xd->lossless);
Jingning Han's avatar
Jingning Han committed
1518
1519
1520
1521
1522
1523
1524
      break;
    default:
      assert(0 && "Invalid transform size");
      break;
  }
}

1525
1526
1527
1528
1529
1530
1531
1532
#if CONFIG_VAR_TX
static void encode_block_inter(int plane, int block, int blk_row, int blk_col,
                               BLOCK_SIZE plane_bsize, TX_SIZE tx_size,
                               void *arg) {
  struct encode_b_args *const args = arg;
  MACROBLOCK *const x = args->x;
  MACROBLOCKD *const xd = &x->e_mbd;
  MB_MODE_INFO *const mbmi = &xd->mi[0]->mbmi;
1533
  const BLOCK_SIZE bsize = txsize_to_bsize[tx_size];
1534
  const struct macroblockd_plane *const pd = &xd->plane[plane];
1535
1536
  int blk_idx = (blk_row >> (1 - pd->subsampling_y)) * 8 +
                (blk_col >> (1 - pd->subsampling_x));
1537
  TX_SIZE plane_tx_size = plane ?
1538
1539
      get_uv_tx_size_impl(mbmi->inter_tx_size[blk_idx], bsize,
                          0, 0) :
1540
      mbmi->inter_tx_size[blk_idx];
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577

  int max_blocks_high = num_4x4_blocks_high_lookup[plane_bsize];
  int max_blocks_wide = num_4x4_blocks_wide_lookup[plane_bsize];

  if (xd->mb_to_bottom_edge < 0)
    max_blocks_high += xd->mb_to_bottom_edge >> (5 + pd->subsampling_y);
  if (xd->mb_to_right_edge < 0)
    max_blocks_wide += xd->mb_to_right_edge >> (5 + pd->subsampling_x);

  if (blk_row >= max_blocks_high || blk_col >= max_blocks_wide)
    return;

  if (tx_size == plane_tx_size) {
    encode_block(plane, block, blk_row, blk_col, plane_bsize,
                 tx_size, arg);
  } else {
    int bsl = b_width_log2_lookup[bsize];
    int i;

    assert(bsl > 0);
    --bsl;

    for (i = 0; i < 4; ++i) {
      const int offsetr = blk_row + ((i >> 1) << bsl);
      const int offsetc = blk_col + ((i & 0x01) << bsl);
      int step = 1 << (2 * (tx_size - 1));

      if (offsetr >= max_blocks_high || offsetc >= max_blocks_wide)
        continue;

      encode_block_inter(plane, block + i * step, offsetr, offsetc,
                         plane_bsize, tx_size - 1, arg);
    }
  }
}
#endif

1578
1579
static void encode_block_pass1(int plane, int block, int blk_row, int blk_col,
                               BLOCK_SIZE plane_bsize,
Jingning Han's avatar
Jingning Han committed
1580
1581
1582
1583
1584
1585
1586
                               TX_SIZE tx_size, void *arg) {
  MACROBLOCK *const x = (MACROBLOCK *)arg;
  MACROBLOCKD *const xd = &x->e_mbd;
  struct macroblock_plane *const p = &x->plane[plane];
  struct macroblockd_plane *const pd = &xd->plane[plane];
  tran_low_t *const dqcoeff = BLOCK_OFFSET(pd->dqcoeff, block);
  uint8_t *dst;
1587
  dst = &pd->dst.buf[4 * blk_row * pd->dst.stride + 4 * blk_col];
Jingning Han's avatar
Jingning Han committed
1588

1589
  vp10_xform_quant(x, plane, block, blk_row, blk_col, plane_bsize, tx_size);
Jingning Han's avatar
Jingning Han committed
1590
1591
1592
1593

  if (p->eobs[block] > 0) {
#if CONFIG_VP9_HIGHBITDEPTH
    if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
1594
1595
1596
1597
1598
1599
1600
1601
      if (xd->lossless) {
        vp10_highbd_iwht4x4_add(dqcoeff, dst, pd->dst.stride,
                                p->eobs[block], xd->bd);
      } else {
        vp10_highbd_idct4x4_add(dqcoeff, dst, pd->dst.stride,
                                p->eobs[block], xd->bd);
      }
      return;
Jingning Han's avatar
Jingning Han committed
1602
1603
    }
#endif  // CONFIG_VP9_HIGHBITDEPTH
1604
1605
1606
1607
1608
    if (xd->lossless) {
      vp10_iwht4x4_add(dqcoeff, dst, pd->dst.stride, p->eobs[block]);
    } else {
      vp10_idct4x4_add(dqcoeff, dst, pd->dst.stride, p->eobs[block]);
    }
Jingning Han's avatar
Jingning Han committed
1609
1610
1611
1612
1613
1614
  }
}

void vp10_encode_sby_pass1(MACROBLOCK *x, BLOCK_SIZE bsize) {
  vp10_subtract_plane(x, bsize, 0);
  vp10_foreach_transformed_block_in_plane(&x->e_mbd, bsize, 0,
1615
                                          encode_block_pass1, x);
Jingning Han's avatar
Jingning Han committed
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
}

void vp10_encode_sb(MACROBLOCK *x, BLOCK_SIZE bsize) {
  MACROBLOCKD *const xd = &x->e_mbd;
  struct optimize_ctx ctx;
  MB_MODE_INFO *mbmi = &xd->mi[0]->mbmi;
  struct encode_b_args arg = {x, &ctx, &mbmi->skip};
  int plane;

  mbmi->skip = 1;

  if (x->skip)
    return;

  for (plane = 0; plane < MAX_MB_PLANE; ++plane) {
1631
1632
1633
1634
1635
1636
1637
#if CONFIG_VAR_TX
    // TODO(jingning): Clean this up.
    const struct macroblockd_plane *const pd = &xd->plane[plane];
    const BLOCK_SIZE plane_bsize = get_plane_block_size(bsize, pd);
    const int mi_width = num_4x4_blocks_wide_lookup[plane_bsize];
    const int mi_height = num_4x4_blocks_high_lookup[plane_bsize];
    const TX_SIZE max_tx_size = max_txsize_lookup[plane_bsize];
1638
1639
    const BLOCK_SIZE txb_size = txsize_to_bsize[max_tx_size];
    const int bh = num_4x4_blocks_wide_lookup[txb_size];
1640
1641
1642
1643
    int idx, idy;
    int block = 0;
    int step = 1 << (max_tx_size * 2);
#endif
Jingning Han's avatar
Jingning Han committed
1644
1645
1646
1647
    if (!x->skip_recode)
      vp10_subtract_plane(x, bsize, plane);

    if (x->optimize && (!x->skip_recode || !x->skip_optimize)) {
1648
1649
1650
1651
#if CONFIG_VAR_TX
      vp10_get_entropy_contexts(bsize, TX_4X4, pd,
                                ctx.ta[plane], ctx.tl[plane]);
#else
Jingning Han's avatar
Jingning Han committed
1652
1653
1654
      const struct macroblockd_plane* const pd = &xd->plane[plane];
      const TX_SIZE tx_size = plane ? get_uv_tx_size(mbmi, pd) : mbmi->tx_size;
      vp10_get_entropy_contexts(bsize, tx_size, pd,
1655
                                ctx.ta[plane], ctx.tl[plane]);
1656
#endif
Jingning Han's avatar
Jingning Han committed
1657
1658
    }

1659
1660
1661
1662
1663
1664
1665
1666
1667
#if CONFIG_VAR_TX
    for (idy = 0; idy < mi_height; idy += bh) {
      for (idx = 0; idx < mi_width; idx += bh) {
        encode_block_inter(plane, block, idy, idx, plane_bsize,
                           max_tx_size, &arg);
        block += step;
      }
    }
#else
Jingning Han's avatar
Jingning Han committed
1668
    vp10_foreach_transformed_block_in_plane(xd, bsize, plane, encode_block,
1669
1670
                                            &arg);
#endif
Jingning Han's avatar
Jingning Han committed
1671
1672
1673
  }
}

1674
1675
1676
void vp10_encode_block_intra(int plane, int block, int blk_row, int blk_col,
                             BLOCK_SIZE plane_bsize,
                             TX_SIZE tx_size, void *arg) {
Jingning Han's avatar
Jingning Han committed
1677
1678
1679
1680
1681
1682
1683
1684
1685
  struct encode_b_args* const args = arg;
  MACROBLOCK *const x = args->x;
  MACROBLOCKD *const xd = &x->e_mbd;
  MB_MODE_INFO *mbmi = &xd->mi[0]->mbmi;
  struct macroblock_plane *const p = &x->plane[plane];
  struct macroblockd_plane *const pd = &xd->plane[plane];
  tran_low_t *coeff = BLOCK_OFFSET(p->coeff, block);
  tran_low_t *qcoeff = BLOCK_OFFSET(p->qcoeff, block);
  tran_low_t *dqcoeff = BLOCK_OFFSET(pd->dqcoeff, block);
hui su's avatar
hui su committed
1686
  PLANE_TYPE plane_type = (plane == 0) ? PLANE_TYPE_Y : PLANE_TYPE_UV;
hui su's avatar
hui su committed
1687
  TX_TYPE tx_type = get_tx_type(plane_type, xd, block, tx_size);
1688
  const scan_order *const scan_order = get_scan(tx_size, tx_type, 0);
Jingning Han's avatar
Jingning Han committed
1689
1690
1691
1692
1693
1694
1695
1696
  PREDICTION_MODE mode;
  const int bwl = b_width_log2_lookup[plane_bsize];
  const int diff_stride = 4 * (1 << bwl);
  uint8_t *src, *dst;
  int16_t *src_diff;
  uint16_t *eob = &p->eobs[block];
  const int src_stride = p->src.stride;
  const int dst_stride = pd->dst.stride;
1697
1698
1699
  dst = &pd->dst.buf[4 * (blk_row * dst_stride + blk_col)];
  src = &p->src.buf[4 * (blk_row * src_stride + blk_col)];
  src_diff = &p->src_diff[4 * (blk_row * diff_stride + blk_col)];
Jingning Han's avatar
Jingning Han committed
1700

hui su's avatar
hui su committed
1701
  mode = plane == 0 ? get_y_mode(xd->mi[0], block) : mbmi->uv_mode;
1702
  vp10_predict_intra_block(xd, bwl, tx_size, mode, dst, dst_stride,
1703
                          dst, dst_stride, blk_col, blk_row, plane);
Jingning Han's avatar
Jingning Han committed
1704
1705
1706
1707
1708
1709
1710
1711

#if CONFIG_VP9_HIGHBITDEPTH
  if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
    switch (tx_size) {
      case TX_32X32:
        if (!x->skip_recode) {
          vpx_highbd_subtract_block(32, 32, src_diff, diff_stride,
                                    src, src_stride,