decodeframe.c 80.8 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
20
21
22
23
24
25
/*
 *  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 <assert.h>
#include <stdlib.h>  // qsort()

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

#include "vpx_dsp/bitreader_buffer.h"
#include "vpx_dsp/bitreader.h"
#include "vpx_mem/vpx_mem.h"
#include "vpx_ports/mem.h"
#include "vpx_ports/mem_ops.h"
#include "vpx_scale/vpx_scale.h"
#include "vpx_util/vpx_thread.h"

26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
#include "vp10/common/alloccommon.h"
#include "vp10/common/common.h"
#include "vp10/common/entropy.h"
#include "vp10/common/entropymode.h"
#include "vp10/common/idct.h"
#include "vp10/common/thread_common.h"
#include "vp10/common/pred_common.h"
#include "vp10/common/quant_common.h"
#include "vp10/common/reconintra.h"
#include "vp10/common/reconinter.h"
#include "vp10/common/seg_common.h"
#include "vp10/common/tile_common.h"

#include "vp10/decoder/decodeframe.h"
#include "vp10/decoder/detokenize.h"
#include "vp10/decoder/decodemv.h"
#include "vp10/decoder/decoder.h"
#include "vp10/decoder/dsubexp.h"
Jingning Han's avatar
Jingning Han committed
44
45
46

#define MAX_VP9_HEADER_SIZE 80

Yaowu Xu's avatar
Yaowu Xu committed
47
static int is_compound_reference_allowed(const VP10_COMMON *cm) {
Jingning Han's avatar
Jingning Han committed
48
49
50
51
52
53
54
55
  int i;
  for (i = 1; i < REFS_PER_FRAME; ++i)
    if (cm->ref_frame_sign_bias[i + 1] != cm->ref_frame_sign_bias[1])
      return 1;

  return 0;
}

Yaowu Xu's avatar
Yaowu Xu committed
56
static void setup_compound_reference_mode(VP10_COMMON *cm) {
Jingning Han's avatar
Jingning Han committed
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
107
108
109
110
111
112
113
114
115
116
117
118
119
  if (cm->ref_frame_sign_bias[LAST_FRAME] ==
          cm->ref_frame_sign_bias[GOLDEN_FRAME]) {
    cm->comp_fixed_ref = ALTREF_FRAME;
    cm->comp_var_ref[0] = LAST_FRAME;
    cm->comp_var_ref[1] = GOLDEN_FRAME;
  } else if (cm->ref_frame_sign_bias[LAST_FRAME] ==
                 cm->ref_frame_sign_bias[ALTREF_FRAME]) {
    cm->comp_fixed_ref = GOLDEN_FRAME;
    cm->comp_var_ref[0] = LAST_FRAME;
    cm->comp_var_ref[1] = ALTREF_FRAME;
  } else {
    cm->comp_fixed_ref = LAST_FRAME;
    cm->comp_var_ref[0] = GOLDEN_FRAME;
    cm->comp_var_ref[1] = ALTREF_FRAME;
  }
}

static int read_is_valid(const uint8_t *start, size_t len, const uint8_t *end) {
  return len != 0 && len <= (size_t)(end - start);
}

static int decode_unsigned_max(struct vpx_read_bit_buffer *rb, int max) {
  const int data = vpx_rb_read_literal(rb, get_unsigned_bits(max));
  return data > max ? max : data;
}

static TX_MODE read_tx_mode(vpx_reader *r) {
  TX_MODE tx_mode = vpx_read_literal(r, 2);
  if (tx_mode == ALLOW_32X32)
    tx_mode += vpx_read_bit(r);
  return tx_mode;
}

static void read_tx_mode_probs(struct tx_probs *tx_probs, vpx_reader *r) {
  int i, j;

  for (i = 0; i < TX_SIZE_CONTEXTS; ++i)
    for (j = 0; j < TX_SIZES - 3; ++j)
      vp10_diff_update_prob(r, &tx_probs->p8x8[i][j]);

  for (i = 0; i < TX_SIZE_CONTEXTS; ++i)
    for (j = 0; j < TX_SIZES - 2; ++j)
      vp10_diff_update_prob(r, &tx_probs->p16x16[i][j]);

  for (i = 0; i < TX_SIZE_CONTEXTS; ++i)
    for (j = 0; j < TX_SIZES - 1; ++j)
      vp10_diff_update_prob(r, &tx_probs->p32x32[i][j]);
}

static void read_switchable_interp_probs(FRAME_CONTEXT *fc, vpx_reader *r) {
  int i, j;
  for (j = 0; j < SWITCHABLE_FILTER_CONTEXTS; ++j)
    for (i = 0; i < SWITCHABLE_FILTERS - 1; ++i)
      vp10_diff_update_prob(r, &fc->switchable_interp_prob[j][i]);
}

static void read_inter_mode_probs(FRAME_CONTEXT *fc, vpx_reader *r) {
  int i, j;
  for (i = 0; i < INTER_MODE_CONTEXTS; ++i)
    for (j = 0; j < INTER_MODES - 1; ++j)
      vp10_diff_update_prob(r, &fc->inter_mode_probs[i][j]);
}

Yaowu Xu's avatar
Yaowu Xu committed
120
static REFERENCE_MODE read_frame_reference_mode(const VP10_COMMON *cm,
Jingning Han's avatar
Jingning Han committed
121
122
123
124
125
126
127
128
129
130
                                                vpx_reader *r) {
  if (is_compound_reference_allowed(cm)) {
    return vpx_read_bit(r) ? (vpx_read_bit(r) ? REFERENCE_MODE_SELECT
                                              : COMPOUND_REFERENCE)
                           : SINGLE_REFERENCE;
  } else {
    return SINGLE_REFERENCE;
  }
}

Yaowu Xu's avatar
Yaowu Xu committed
131
static void read_frame_reference_mode_probs(VP10_COMMON *cm, vpx_reader *r) {
Jingning Han's avatar
Jingning Han committed
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
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
  FRAME_CONTEXT *const fc = cm->fc;
  int i;

  if (cm->reference_mode == REFERENCE_MODE_SELECT)
    for (i = 0; i < COMP_INTER_CONTEXTS; ++i)
      vp10_diff_update_prob(r, &fc->comp_inter_prob[i]);

  if (cm->reference_mode != COMPOUND_REFERENCE)
    for (i = 0; i < REF_CONTEXTS; ++i) {
      vp10_diff_update_prob(r, &fc->single_ref_prob[i][0]);
      vp10_diff_update_prob(r, &fc->single_ref_prob[i][1]);
    }

  if (cm->reference_mode != SINGLE_REFERENCE)
    for (i = 0; i < REF_CONTEXTS; ++i)
      vp10_diff_update_prob(r, &fc->comp_ref_prob[i]);
}

static void update_mv_probs(vpx_prob *p, int n, vpx_reader *r) {
  int i;
  for (i = 0; i < n; ++i)
    if (vpx_read(r, MV_UPDATE_PROB))
      p[i] = (vpx_read_literal(r, 7) << 1) | 1;
}

static void read_mv_probs(nmv_context *ctx, int allow_hp, vpx_reader *r) {
  int i, j;

  update_mv_probs(ctx->joints, MV_JOINTS - 1, r);

  for (i = 0; i < 2; ++i) {
    nmv_component *const comp_ctx = &ctx->comps[i];
    update_mv_probs(&comp_ctx->sign, 1, r);
    update_mv_probs(comp_ctx->classes, MV_CLASSES - 1, r);
    update_mv_probs(comp_ctx->class0, CLASS0_SIZE - 1, r);
    update_mv_probs(comp_ctx->bits, MV_OFFSET_BITS, r);
  }

  for (i = 0; i < 2; ++i) {
    nmv_component *const comp_ctx = &ctx->comps[i];
    for (j = 0; j < CLASS0_SIZE; ++j)
      update_mv_probs(comp_ctx->class0_fp[j], MV_FP_SIZE - 1, r);
    update_mv_probs(comp_ctx->fp, 3, r);
  }

  if (allow_hp) {
    for (i = 0; i < 2; ++i) {
      nmv_component *const comp_ctx = &ctx->comps[i];
      update_mv_probs(&comp_ctx->class0_hp, 1, r);
      update_mv_probs(&comp_ctx->hp, 1, r);
    }
  }
}

static void inverse_transform_block_inter(MACROBLOCKD* xd, int plane,
                                          const TX_SIZE tx_size,
                                          uint8_t *dst, int stride,
                                          int eob) {
  struct macroblockd_plane *const pd = &xd->plane[plane];
  if (eob > 0) {
    tran_low_t *const dqcoeff = pd->dqcoeff;
#if CONFIG_VP9_HIGHBITDEPTH
    if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
      if (xd->lossless) {
        vp10_highbd_iwht4x4_add(dqcoeff, dst, stride, eob, xd->bd);
      } else {
        switch (tx_size) {
          case TX_4X4:
            vp10_highbd_idct4x4_add(dqcoeff, dst, stride, eob, xd->bd);
            break;
          case TX_8X8:
            vp10_highbd_idct8x8_add(dqcoeff, dst, stride, eob, xd->bd);
            break;
          case TX_16X16:
            vp10_highbd_idct16x16_add(dqcoeff, dst, stride, eob, xd->bd);
            break;
          case TX_32X32:
            vp10_highbd_idct32x32_add(dqcoeff, dst, stride, eob, xd->bd);
            break;
          default:
            assert(0 && "Invalid transform size");
        }
      }
    } else {
      if (xd->lossless) {
        vp10_iwht4x4_add(dqcoeff, dst, stride, eob);
      } else {
        switch (tx_size) {
          case TX_4X4:
            vp10_idct4x4_add(dqcoeff, dst, stride, eob);
            break;
          case TX_8X8:
            vp10_idct8x8_add(dqcoeff, dst, stride, eob);
            break;
          case TX_16X16:
            vp10_idct16x16_add(dqcoeff, dst, stride, eob);
            break;
          case TX_32X32:
            vp10_idct32x32_add(dqcoeff, dst, stride, eob);
            break;
          default:
            assert(0 && "Invalid transform size");
            return;
        }
      }
    }
#else
    if (xd->lossless) {
      vp10_iwht4x4_add(dqcoeff, dst, stride, eob);
    } else {
      switch (tx_size) {
        case TX_4X4:
          vp10_idct4x4_add(dqcoeff, dst, stride, eob);
          break;
        case TX_8X8:
          vp10_idct8x8_add(dqcoeff, dst, stride, eob);
          break;
        case TX_16X16:
          vp10_idct16x16_add(dqcoeff, dst, stride, eob);
          break;
        case TX_32X32:
          vp10_idct32x32_add(dqcoeff, dst, stride, eob);
          break;
        default:
          assert(0 && "Invalid transform size");
          return;
      }
    }
#endif  // CONFIG_VP9_HIGHBITDEPTH

    if (eob == 1) {
      dqcoeff[0] = 0;
    } else {
      if (tx_size <= TX_16X16 && eob <= 10)
        memset(dqcoeff, 0, 4 * (4 << tx_size) * sizeof(dqcoeff[0]));
      else if (tx_size == TX_32X32 && eob <= 34)
        memset(dqcoeff, 0, 256 * sizeof(dqcoeff[0]));
      else
        memset(dqcoeff, 0, (16 << (tx_size << 1)) * sizeof(dqcoeff[0]));
    }
  }
}

static void inverse_transform_block_intra(MACROBLOCKD* xd, int plane,
                                          const TX_TYPE tx_type,
                                          const TX_SIZE tx_size,
                                          uint8_t *dst, int stride,
                                          int eob) {
  struct macroblockd_plane *const pd = &xd->plane[plane];
  if (eob > 0) {
    tran_low_t *const dqcoeff = pd->dqcoeff;
#if CONFIG_VP9_HIGHBITDEPTH
    if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
      if (xd->lossless) {
        vp10_highbd_iwht4x4_add(dqcoeff, dst, stride, eob, xd->bd);
      } else {
        switch (tx_size) {
          case TX_4X4:
            vp10_highbd_iht4x4_add(tx_type, dqcoeff, dst, stride, eob, xd->bd);
            break;
          case TX_8X8:
            vp10_highbd_iht8x8_add(tx_type, dqcoeff, dst, stride, eob, xd->bd);
            break;
          case TX_16X16:
            vp10_highbd_iht16x16_add(tx_type, dqcoeff, dst, stride, eob, xd->bd);
            break;
          case TX_32X32:
            vp10_highbd_idct32x32_add(dqcoeff, dst, stride, eob, xd->bd);
            break;
          default:
            assert(0 && "Invalid transform size");
        }
      }
    } else {
      if (xd->lossless) {
        vp10_iwht4x4_add(dqcoeff, dst, stride, eob);
      } else {
        switch (tx_size) {
          case TX_4X4:
            vp10_iht4x4_add(tx_type, dqcoeff, dst, stride, eob);
            break;
          case TX_8X8:
            vp10_iht8x8_add(tx_type, dqcoeff, dst, stride, eob);
            break;
          case TX_16X16:
            vp10_iht16x16_add(tx_type, dqcoeff, dst, stride, eob);
            break;
          case TX_32X32:
            vp10_idct32x32_add(dqcoeff, dst, stride, eob);
            break;
          default:
            assert(0 && "Invalid transform size");
            return;
        }
      }
    }
#else
    if (xd->lossless) {
      vp10_iwht4x4_add(dqcoeff, dst, stride, eob);
    } else {
      switch (tx_size) {
        case TX_4X4:
          vp10_iht4x4_add(tx_type, dqcoeff, dst, stride, eob);
          break;
        case TX_8X8:
          vp10_iht8x8_add(tx_type, dqcoeff, dst, stride, eob);
          break;
        case TX_16X16:
          vp10_iht16x16_add(tx_type, dqcoeff, dst, stride, eob);
          break;
        case TX_32X32:
          vp10_idct32x32_add(dqcoeff, dst, stride, eob);
          break;
        default:
          assert(0 && "Invalid transform size");
          return;
      }
    }
#endif  // CONFIG_VP9_HIGHBITDEPTH

    if (eob == 1) {
      dqcoeff[0] = 0;
    } else {
      if (tx_type == DCT_DCT && tx_size <= TX_16X16 && eob <= 10)
        memset(dqcoeff, 0, 4 * (4 << tx_size) * sizeof(dqcoeff[0]));
      else if (tx_size == TX_32X32 && eob <= 34)
        memset(dqcoeff, 0, 256 * sizeof(dqcoeff[0]));
      else
        memset(dqcoeff, 0, (16 << (tx_size << 1)) * sizeof(dqcoeff[0]));
    }
  }
}

static void predict_and_reconstruct_intra_block(MACROBLOCKD *const xd,
                                                vpx_reader *r,
                                                MB_MODE_INFO *const mbmi,
                                                int plane,
                                                int row, int col,
                                                TX_SIZE tx_size) {
  struct macroblockd_plane *const pd = &xd->plane[plane];
  PREDICTION_MODE mode = (plane == 0) ? mbmi->mode : mbmi->uv_mode;
  uint8_t *dst;
  dst = &pd->dst.buf[4 * row * pd->dst.stride + 4 * col];

  if (mbmi->sb_type < BLOCK_8X8)
    if (plane == 0)
      mode = xd->mi[0]->bmi[(row << 1) + col].as_mode;

  vp10_predict_intra_block(xd, pd->n4_wl, tx_size, mode,
                          dst, pd->dst.stride, dst, pd->dst.stride,
                          col, row, plane);

  if (!mbmi->skip) {
    const TX_TYPE tx_type = (plane || xd->lossless) ?
        DCT_DCT : intra_mode_to_tx_type_lookup[mode];
    const scan_order *sc = (plane || xd->lossless) ?
        &vp10_default_scan_orders[tx_size] : &vp10_scan_orders[tx_size][tx_type];
    const int eob = vp10_decode_block_tokens(xd, plane, sc, col, row, tx_size,
                                            r, mbmi->segment_id);
    inverse_transform_block_intra(xd, plane, tx_type, tx_size,
                                  dst, pd->dst.stride, eob);
  }
}

static int reconstruct_inter_block(MACROBLOCKD *const xd, vpx_reader *r,
                                   MB_MODE_INFO *const mbmi, int plane,
                                   int row, int col, TX_SIZE tx_size) {
  struct macroblockd_plane *const pd = &xd->plane[plane];
  const scan_order *sc = &vp10_default_scan_orders[tx_size];
  const int eob = vp10_decode_block_tokens(xd, plane, sc, col, row, tx_size, r,
                                          mbmi->segment_id);

  inverse_transform_block_inter(xd, plane, tx_size,
                            &pd->dst.buf[4 * row * pd->dst.stride + 4 * col],
                            pd->dst.stride, eob);
  return eob;
}

static void build_mc_border(const uint8_t *src, int src_stride,
                            uint8_t *dst, int dst_stride,
                            int x, int y, int b_w, int b_h, int w, int h) {
  // Get a pointer to the start of the real data for this row.
  const uint8_t *ref_row = src - x - y * src_stride;

  if (y >= h)
    ref_row += (h - 1) * src_stride;
  else if (y > 0)
    ref_row += y * src_stride;

  do {
    int right = 0, copy;
    int left = x < 0 ? -x : 0;

    if (left > b_w)
      left = b_w;

    if (x + b_w > w)
      right = x + b_w - w;

    if (right > b_w)
      right = b_w;

    copy = b_w - left - right;

    if (left)
      memset(dst, ref_row[0], left);

    if (copy)
      memcpy(dst + left, ref_row + x + left, copy);

    if (right)
      memset(dst + left + copy, ref_row[w - 1], right);

    dst += dst_stride;
    ++y;

    if (y > 0 && y < h)
      ref_row += src_stride;
  } while (--b_h);
}

#if CONFIG_VP9_HIGHBITDEPTH
static void high_build_mc_border(const uint8_t *src8, int src_stride,
                                 uint16_t *dst, int dst_stride,
                                 int x, int y, int b_w, int b_h,
                                 int w, int h) {
  // Get a pointer to the start of the real data for this row.
  const uint16_t *src = CONVERT_TO_SHORTPTR(src8);
  const uint16_t *ref_row = src - x - y * src_stride;

  if (y >= h)
    ref_row += (h - 1) * src_stride;
  else if (y > 0)
    ref_row += y * src_stride;

  do {
    int right = 0, copy;
    int left = x < 0 ? -x : 0;

    if (left > b_w)
      left = b_w;

    if (x + b_w > w)
      right = x + b_w - w;

    if (right > b_w)
      right = b_w;

    copy = b_w - left - right;

    if (left)
      vpx_memset16(dst, ref_row[0], left);

    if (copy)
      memcpy(dst + left, ref_row + x + left, copy * sizeof(uint16_t));

    if (right)
      vpx_memset16(dst + left + copy, ref_row[w - 1], right);

    dst += dst_stride;
    ++y;

    if (y > 0 && y < h)
      ref_row += src_stride;
  } while (--b_h);
}
#endif  // CONFIG_VP9_HIGHBITDEPTH

#if CONFIG_VP9_HIGHBITDEPTH
static void extend_and_predict(const uint8_t *buf_ptr1, int pre_buf_stride,
                               int x0, int y0, int b_w, int b_h,
                               int frame_width, int frame_height,
                               int border_offset,
                               uint8_t *const dst, int dst_buf_stride,
                               int subpel_x, int subpel_y,
                               const InterpKernel *kernel,
                               const struct scale_factors *sf,
                               MACROBLOCKD *xd,
                               int w, int h, int ref, int xs, int ys) {
  DECLARE_ALIGNED(16, uint16_t, mc_buf_high[80 * 2 * 80 * 2]);
  const uint8_t *buf_ptr;

  if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
    high_build_mc_border(buf_ptr1, pre_buf_stride, mc_buf_high, b_w,
                         x0, y0, b_w, b_h, frame_width, frame_height);
    buf_ptr = CONVERT_TO_BYTEPTR(mc_buf_high) + border_offset;
  } else {
    build_mc_border(buf_ptr1, pre_buf_stride, (uint8_t *)mc_buf_high, b_w,
                    x0, y0, b_w, b_h, frame_width, frame_height);
    buf_ptr = ((uint8_t *)mc_buf_high) + border_offset;
  }

  if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
    high_inter_predictor(buf_ptr, b_w, dst, dst_buf_stride, subpel_x,
                         subpel_y, sf, w, h, ref, kernel, xs, ys, xd->bd);
  } else {
    inter_predictor(buf_ptr, b_w, dst, dst_buf_stride, subpel_x,
                    subpel_y, sf, w, h, ref, kernel, xs, ys);
  }
}
#else
static void extend_and_predict(const uint8_t *buf_ptr1, int pre_buf_stride,
                               int x0, int y0, int b_w, int b_h,
                               int frame_width, int frame_height,
                               int border_offset,
                               uint8_t *const dst, int dst_buf_stride,
                               int subpel_x, int subpel_y,
                               const InterpKernel *kernel,
                               const struct scale_factors *sf,
                               int w, int h, int ref, int xs, int ys) {
  DECLARE_ALIGNED(16, uint8_t, mc_buf[80 * 2 * 80 * 2]);
  const uint8_t *buf_ptr;

  build_mc_border(buf_ptr1, pre_buf_stride, mc_buf, b_w,
                  x0, y0, b_w, b_h, frame_width, frame_height);
  buf_ptr = mc_buf + border_offset;

  inter_predictor(buf_ptr, b_w, dst, dst_buf_stride, subpel_x,
                  subpel_y, sf, w, h, ref, kernel, xs, ys);
}
#endif  // CONFIG_VP9_HIGHBITDEPTH

Yaowu Xu's avatar
Yaowu Xu committed
554
static void dec_build_inter_predictors(VP10Decoder *const pbi, MACROBLOCKD *xd,
Jingning Han's avatar
Jingning Han committed
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
                                       int plane, int bw, int bh, int x,
                                       int y, int w, int h, int mi_x, int mi_y,
                                       const InterpKernel *kernel,
                                       const struct scale_factors *sf,
                                       struct buf_2d *pre_buf,
                                       struct buf_2d *dst_buf, const MV* mv,
                                       RefCntBuffer *ref_frame_buf,
                                       int is_scaled, int ref) {
  struct macroblockd_plane *const pd = &xd->plane[plane];
  uint8_t *const dst = dst_buf->buf + dst_buf->stride * y + x;
  MV32 scaled_mv;
  int xs, ys, x0, y0, x0_16, y0_16, frame_width, frame_height,
      buf_stride, subpel_x, subpel_y;
  uint8_t *ref_frame, *buf_ptr;

  // Get reference frame pointer, width and height.
  if (plane == 0) {
    frame_width = ref_frame_buf->buf.y_crop_width;
    frame_height = ref_frame_buf->buf.y_crop_height;
    ref_frame = ref_frame_buf->buf.y_buffer;
  } else {
    frame_width = ref_frame_buf->buf.uv_crop_width;
    frame_height = ref_frame_buf->buf.uv_crop_height;
    ref_frame = plane == 1 ? ref_frame_buf->buf.u_buffer
                         : ref_frame_buf->buf.v_buffer;
  }

  if (is_scaled) {
    const MV mv_q4 = clamp_mv_to_umv_border_sb(xd, mv, bw, bh,
                                               pd->subsampling_x,
                                               pd->subsampling_y);
    // Co-ordinate of containing block to pixel precision.
    int x_start = (-xd->mb_to_left_edge >> (3 + pd->subsampling_x));
    int y_start = (-xd->mb_to_top_edge >> (3 + pd->subsampling_y));

    // Co-ordinate of the block to 1/16th pixel precision.
    x0_16 = (x_start + x) << SUBPEL_BITS;
    y0_16 = (y_start + y) << SUBPEL_BITS;

    // Co-ordinate of current block in reference frame
    // to 1/16th pixel precision.
    x0_16 = sf->scale_value_x(x0_16, sf);
    y0_16 = sf->scale_value_y(y0_16, sf);

    // Map the top left corner of the block into the reference frame.
    x0 = sf->scale_value_x(x_start + x, sf);
    y0 = sf->scale_value_y(y_start + y, sf);

    // Scale the MV and incorporate the sub-pixel offset of the block
    // in the reference frame.
    scaled_mv = vp10_scale_mv(&mv_q4, mi_x + x, mi_y + y, sf);
    xs = sf->x_step_q4;
    ys = sf->y_step_q4;
  } else {
    // Co-ordinate of containing block to pixel precision.
    x0 = (-xd->mb_to_left_edge >> (3 + pd->subsampling_x)) + x;
    y0 = (-xd->mb_to_top_edge >> (3 + pd->subsampling_y)) + y;

    // Co-ordinate of the block to 1/16th pixel precision.
    x0_16 = x0 << SUBPEL_BITS;
    y0_16 = y0 << SUBPEL_BITS;

    scaled_mv.row = mv->row * (1 << (1 - pd->subsampling_y));
    scaled_mv.col = mv->col * (1 << (1 - pd->subsampling_x));
    xs = ys = 16;
  }
  subpel_x = scaled_mv.col & SUBPEL_MASK;
  subpel_y = scaled_mv.row & SUBPEL_MASK;

  // Calculate the top left corner of the best matching block in the
  // reference frame.
  x0 += scaled_mv.col >> SUBPEL_BITS;
  y0 += scaled_mv.row >> SUBPEL_BITS;
  x0_16 += scaled_mv.col;
  y0_16 += scaled_mv.row;

  // Get reference block pointer.
  buf_ptr = ref_frame + y0 * pre_buf->stride + x0;
  buf_stride = pre_buf->stride;

  // Do border extension if there is motion or the
  // width/height is not a multiple of 8 pixels.
  if (is_scaled || scaled_mv.col || scaled_mv.row ||
      (frame_width & 0x7) || (frame_height & 0x7)) {
    int y1 = ((y0_16 + (h - 1) * ys) >> SUBPEL_BITS) + 1;

    // Get reference block bottom right horizontal coordinate.
    int x1 = ((x0_16 + (w - 1) * xs) >> SUBPEL_BITS) + 1;
    int x_pad = 0, y_pad = 0;

    if (subpel_x || (sf->x_step_q4 != SUBPEL_SHIFTS)) {
      x0 -= VP9_INTERP_EXTEND - 1;
      x1 += VP9_INTERP_EXTEND;
      x_pad = 1;
    }

    if (subpel_y || (sf->y_step_q4 != SUBPEL_SHIFTS)) {
      y0 -= VP9_INTERP_EXTEND - 1;
      y1 += VP9_INTERP_EXTEND;
      y_pad = 1;
    }

    // Wait until reference block is ready. Pad 7 more pixels as last 7
    // pixels of each superblock row can be changed by next superblock row.
    if (pbi->frame_parallel_decode)
      vp10_frameworker_wait(pbi->frame_worker_owner, ref_frame_buf,
                           MAX(0, (y1 + 7)) << (plane == 0 ? 0 : 1));

    // Skip border extension if block is inside the frame.
    if (x0 < 0 || x0 > frame_width - 1 || x1 < 0 || x1 > frame_width - 1 ||
        y0 < 0 || y0 > frame_height - 1 || y1 < 0 || y1 > frame_height - 1) {
      // Extend the border.
      const uint8_t *const buf_ptr1 = ref_frame + y0 * buf_stride + x0;
      const int b_w = x1 - x0 + 1;
      const int b_h = y1 - y0 + 1;
      const int border_offset = y_pad * 3 * b_w + x_pad * 3;

      extend_and_predict(buf_ptr1, buf_stride, x0, y0, b_w, b_h,
                         frame_width, frame_height, border_offset,
                         dst, dst_buf->stride,
                         subpel_x, subpel_y,
                         kernel, sf,
#if CONFIG_VP9_HIGHBITDEPTH
                         xd,
#endif
                         w, h, ref, xs, ys);
      return;
    }
  } else {
    // Wait until reference block is ready. Pad 7 more pixels as last 7
    // pixels of each superblock row can be changed by next superblock row.
     if (pbi->frame_parallel_decode) {
       const int y1 = (y0_16 + (h - 1) * ys) >> SUBPEL_BITS;
       vp10_frameworker_wait(pbi->frame_worker_owner, ref_frame_buf,
                            MAX(0, (y1 + 7)) << (plane == 0 ? 0 : 1));
     }
  }
#if CONFIG_VP9_HIGHBITDEPTH
  if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH) {
    high_inter_predictor(buf_ptr, buf_stride, dst, dst_buf->stride, subpel_x,
                         subpel_y, sf, w, h, ref, kernel, xs, ys, xd->bd);
  } else {
    inter_predictor(buf_ptr, buf_stride, dst, dst_buf->stride, subpel_x,
                    subpel_y, sf, w, h, ref, kernel, xs, ys);
  }
#else
  inter_predictor(buf_ptr, buf_stride, dst, dst_buf->stride, subpel_x,
                  subpel_y, sf, w, h, ref, kernel, xs, ys);
#endif  // CONFIG_VP9_HIGHBITDEPTH
}

Yaowu Xu's avatar
Yaowu Xu committed
706
static void dec_build_inter_predictors_sb(VP10Decoder *const pbi,
Jingning Han's avatar
Jingning Han committed
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
                                          MACROBLOCKD *xd,
                                          int mi_row, int mi_col) {
  int plane;
  const int mi_x = mi_col * MI_SIZE;
  const int mi_y = mi_row * MI_SIZE;
  const MODE_INFO *mi = xd->mi[0];
  const InterpKernel *kernel = vp10_filter_kernels[mi->mbmi.interp_filter];
  const BLOCK_SIZE sb_type = mi->mbmi.sb_type;
  const int is_compound = has_second_ref(&mi->mbmi);

  for (plane = 0; plane < MAX_MB_PLANE; ++plane) {
    struct macroblockd_plane *const pd = &xd->plane[plane];
    struct buf_2d *const dst_buf = &pd->dst;
    const int num_4x4_w = pd->n4_w;
    const int num_4x4_h = pd->n4_h;

    const int n4w_x4 = 4 * num_4x4_w;
    const int n4h_x4 = 4 * num_4x4_h;
    int ref;

    for (ref = 0; ref < 1 + is_compound; ++ref) {
      const struct scale_factors *const sf = &xd->block_refs[ref]->sf;
      struct buf_2d *const pre_buf = &pd->pre[ref];
      const int idx = xd->block_refs[ref]->idx;
      BufferPool *const pool = pbi->common.buffer_pool;
      RefCntBuffer *const ref_frame_buf = &pool->frame_bufs[idx];
      const int is_scaled = vp10_is_scaled(sf);

      if (sb_type < BLOCK_8X8) {
        int i = 0, x, y;
        for (y = 0; y < num_4x4_h; ++y) {
          for (x = 0; x < num_4x4_w; ++x) {
            const MV mv = average_split_mvs(pd, mi, ref, i++);
            dec_build_inter_predictors(pbi, xd, plane, n4w_x4, n4h_x4,
                                       4 * x, 4 * y, 4, 4, mi_x, mi_y, kernel,
                                       sf, pre_buf, dst_buf, &mv,
                                       ref_frame_buf, is_scaled, ref);
          }
        }
      } else {
        const MV mv = mi->mbmi.mv[ref].as_mv;
        dec_build_inter_predictors(pbi, xd, plane, n4w_x4, n4h_x4,
                                   0, 0, n4w_x4, n4h_x4, mi_x, mi_y, kernel,
                                   sf, pre_buf, dst_buf, &mv, ref_frame_buf,
                                   is_scaled, ref);
      }
    }
  }
}

static INLINE TX_SIZE dec_get_uv_tx_size(const MB_MODE_INFO *mbmi,
                                         int n4_wl, int n4_hl) {
  // get minimum log2 num4x4s dimension
  const int x = MIN(n4_wl, n4_hl);
  return MIN(mbmi->tx_size,  x);
}

static INLINE void dec_reset_skip_context(MACROBLOCKD *xd) {
  int i;
  for (i = 0; i < MAX_MB_PLANE; i++) {
    struct macroblockd_plane *const pd = &xd->plane[i];
    memset(pd->above_context, 0, sizeof(ENTROPY_CONTEXT) * pd->n4_w);
    memset(pd->left_context, 0, sizeof(ENTROPY_CONTEXT) * pd->n4_h);
  }
}

static void set_plane_n4(MACROBLOCKD *const xd, int bw, int bh, int bwl,
                         int bhl) {
  int i;
  for (i = 0; i < MAX_MB_PLANE; i++) {
    xd->plane[i].n4_w = (bw << 1) >> xd->plane[i].subsampling_x;
    xd->plane[i].n4_h = (bh << 1) >> xd->plane[i].subsampling_y;
    xd->plane[i].n4_wl = bwl - xd->plane[i].subsampling_x;
    xd->plane[i].n4_hl = bhl - xd->plane[i].subsampling_y;
  }
}

Yaowu Xu's avatar
Yaowu Xu committed
784
static MB_MODE_INFO *set_offsets(VP10_COMMON *const cm, MACROBLOCKD *const xd,
Jingning Han's avatar
Jingning Han committed
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
                                 BLOCK_SIZE bsize, int mi_row, int mi_col,
                                 int bw, int bh, int x_mis, int y_mis,
                                 int bwl, int bhl) {
  const int offset = mi_row * cm->mi_stride + mi_col;
  int x, y;
  const TileInfo *const tile = &xd->tile;

  xd->mi = cm->mi_grid_visible + offset;
  xd->mi[0] = &cm->mi[offset];
  // TODO(slavarnway): Generate sb_type based on bwl and bhl, instead of
  // passing bsize from decode_partition().
  xd->mi[0]->mbmi.sb_type = bsize;
  for (y = 0; y < y_mis; ++y)
    for (x = !y; x < x_mis; ++x) {
      xd->mi[y * cm->mi_stride + x] = xd->mi[0];
    }

  set_plane_n4(xd, bw, bh, bwl, bhl);

  set_skip_context(xd, mi_row, mi_col);

  // Distance of Mb to the various image edges. These are specified to 8th pel
  // as they are always compared to values that are in 1/8th pel units
  set_mi_row_col(xd, tile, mi_row, bh, mi_col, bw, cm->mi_rows, cm->mi_cols);

  vp10_setup_dst_planes(xd->plane, get_frame_new_buffer(cm), mi_row, mi_col);
  return &xd->mi[0]->mbmi;
}

Yaowu Xu's avatar
Yaowu Xu committed
814
static void decode_block(VP10Decoder *const pbi, MACROBLOCKD *const xd,
Jingning Han's avatar
Jingning Han committed
815
816
817
                         int mi_row, int mi_col,
                         vpx_reader *r, BLOCK_SIZE bsize,
                         int bwl, int bhl) {
Yaowu Xu's avatar
Yaowu Xu committed
818
  VP10_COMMON *const cm = &pbi->common;
Jingning Han's avatar
Jingning Han committed
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
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
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
  const int less8x8 = bsize < BLOCK_8X8;
  const int bw = 1 << (bwl - 1);
  const int bh = 1 << (bhl - 1);
  const int x_mis = MIN(bw, cm->mi_cols - mi_col);
  const int y_mis = MIN(bh, cm->mi_rows - mi_row);

  MB_MODE_INFO *mbmi = set_offsets(cm, xd, bsize, mi_row, mi_col,
                                   bw, bh, x_mis, y_mis, bwl, bhl);

  if (bsize >= BLOCK_8X8 && (cm->subsampling_x || cm->subsampling_y)) {
    const BLOCK_SIZE uv_subsize =
        ss_size_lookup[bsize][cm->subsampling_x][cm->subsampling_y];
    if (uv_subsize == BLOCK_INVALID)
      vpx_internal_error(xd->error_info,
                         VPX_CODEC_CORRUPT_FRAME, "Invalid block size.");
  }

  vp10_read_mode_info(pbi, xd, mi_row, mi_col, r, x_mis, y_mis);

  if (mbmi->skip) {
    dec_reset_skip_context(xd);
  }

  if (!is_inter_block(mbmi)) {
    int plane;
    for (plane = 0; plane < MAX_MB_PLANE; ++plane) {
      const struct macroblockd_plane *const pd = &xd->plane[plane];
      const TX_SIZE tx_size =
          plane ? dec_get_uv_tx_size(mbmi, pd->n4_wl, pd->n4_hl)
                  : mbmi->tx_size;
      const int num_4x4_w = pd->n4_w;
      const int num_4x4_h = pd->n4_h;
      const int step = (1 << tx_size);
      int row, col;
      const int max_blocks_wide = num_4x4_w + (xd->mb_to_right_edge >= 0 ?
          0 : xd->mb_to_right_edge >> (5 + pd->subsampling_x));
      const int max_blocks_high = num_4x4_h + (xd->mb_to_bottom_edge >= 0 ?
          0 : xd->mb_to_bottom_edge >> (5 + pd->subsampling_y));

      for (row = 0; row < max_blocks_high; row += step)
        for (col = 0; col < max_blocks_wide; col += step)
          predict_and_reconstruct_intra_block(xd, r, mbmi, plane,
                                              row, col, tx_size);
    }
  } else {
    // Prediction
    dec_build_inter_predictors_sb(pbi, xd, mi_row, mi_col);

    // Reconstruction
    if (!mbmi->skip) {
      int eobtotal = 0;
      int plane;

      for (plane = 0; plane < MAX_MB_PLANE; ++plane) {
        const struct macroblockd_plane *const pd = &xd->plane[plane];
        const TX_SIZE tx_size =
            plane ? dec_get_uv_tx_size(mbmi, pd->n4_wl, pd->n4_hl)
                    : mbmi->tx_size;
        const int num_4x4_w = pd->n4_w;
        const int num_4x4_h = pd->n4_h;
        const int step = (1 << tx_size);
        int row, col;
        const int max_blocks_wide = num_4x4_w + (xd->mb_to_right_edge >= 0 ?
            0 : xd->mb_to_right_edge >> (5 + pd->subsampling_x));
        const int max_blocks_high = num_4x4_h + (xd->mb_to_bottom_edge >= 0 ?
            0 : xd->mb_to_bottom_edge >> (5 + pd->subsampling_y));

        for (row = 0; row < max_blocks_high; row += step)
          for (col = 0; col < max_blocks_wide; col += step)
            eobtotal += reconstruct_inter_block(xd, r, mbmi, plane, row, col,
                                                tx_size);
      }

      if (!less8x8 && eobtotal == 0)
        mbmi->skip = 1;  // skip loopfilter
    }
  }

  xd->corrupted |= vpx_reader_has_error(r);
}

static INLINE int dec_partition_plane_context(const MACROBLOCKD *xd,
                                              int mi_row, int mi_col,
                                              int bsl) {
  const PARTITION_CONTEXT *above_ctx = xd->above_seg_context + mi_col;
  const PARTITION_CONTEXT *left_ctx = xd->left_seg_context + (mi_row & MI_MASK);
  int above = (*above_ctx >> bsl) & 1 , left = (*left_ctx >> bsl) & 1;

//  assert(bsl >= 0);

  return (left * 2 + above) + bsl * PARTITION_PLOFFSET;
}

static INLINE void dec_update_partition_context(MACROBLOCKD *xd,
                                                int mi_row, int mi_col,
                                                BLOCK_SIZE subsize,
                                                int bw) {
  PARTITION_CONTEXT *const above_ctx = xd->above_seg_context + mi_col;
  PARTITION_CONTEXT *const left_ctx = xd->left_seg_context + (mi_row & MI_MASK);

  // update the partition context at the end notes. set partition bits
  // of block sizes larger than the current one to be one, and partition
  // bits of smaller block sizes to be zero.
  memset(above_ctx, partition_context_lookup[subsize].above, bw);
  memset(left_ctx, partition_context_lookup[subsize].left, bw);
}

static PARTITION_TYPE read_partition(MACROBLOCKD *xd, int mi_row, int mi_col,
                                     vpx_reader *r,
                                     int has_rows, int has_cols, int bsl) {
  const int ctx = dec_partition_plane_context(xd, mi_row, mi_col, bsl);
  const vpx_prob *const probs = get_partition_probs(xd, ctx);
  FRAME_COUNTS *counts = xd->counts;
  PARTITION_TYPE p;

  if (has_rows && has_cols)
    p = (PARTITION_TYPE)vpx_read_tree(r, vp10_partition_tree, probs);
  else if (!has_rows && has_cols)
    p = vpx_read(r, probs[1]) ? PARTITION_SPLIT : PARTITION_HORZ;
  else if (has_rows && !has_cols)
    p = vpx_read(r, probs[2]) ? PARTITION_SPLIT : PARTITION_VERT;
  else
    p = PARTITION_SPLIT;

  if (counts)
    ++counts->partition[ctx][p];

  return p;
}

// TODO(slavarnway): eliminate bsize and subsize in future commits
Yaowu Xu's avatar
Yaowu Xu committed
950
static void decode_partition(VP10Decoder *const pbi, MACROBLOCKD *const xd,
Jingning Han's avatar
Jingning Han committed
951
952
                             int mi_row, int mi_col,
                             vpx_reader* r, BLOCK_SIZE bsize, int n4x4_l2) {
Yaowu Xu's avatar
Yaowu Xu committed
953
  VP10_COMMON *const cm = &pbi->common;
Jingning Han's avatar
Jingning Han committed
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
  const int n8x8_l2 = n4x4_l2 - 1;
  const int num_8x8_wh = 1 << n8x8_l2;
  const int hbs = num_8x8_wh >> 1;
  PARTITION_TYPE partition;
  BLOCK_SIZE subsize;
  const int has_rows = (mi_row + hbs) < cm->mi_rows;
  const int has_cols = (mi_col + hbs) < cm->mi_cols;

  if (mi_row >= cm->mi_rows || mi_col >= cm->mi_cols)
    return;

  partition = read_partition(xd, mi_row, mi_col, r, has_rows, has_cols,
                             n8x8_l2);
  subsize = subsize_lookup[partition][bsize];  // get_subsize(bsize, partition);
  if (!hbs) {
    // calculate bmode block dimensions (log 2)
    xd->bmode_blocks_wl = 1 >> !!(partition & PARTITION_VERT);
    xd->bmode_blocks_hl = 1 >> !!(partition & PARTITION_HORZ);
    decode_block(pbi, xd, mi_row, mi_col, r, subsize, 1, 1);
  } else {
    switch (partition) {
      case PARTITION_NONE:
        decode_block(pbi, xd, mi_row, mi_col, r, subsize, n4x4_l2, n4x4_l2);
        break;
      case PARTITION_HORZ:
        decode_block(pbi, xd, mi_row, mi_col, r, subsize, n4x4_l2, n8x8_l2);
        if (has_rows)
          decode_block(pbi, xd, mi_row + hbs, mi_col, r, subsize, n4x4_l2,
                       n8x8_l2);
        break;
      case PARTITION_VERT:
        decode_block(pbi, xd, mi_row, mi_col, r, subsize, n8x8_l2, n4x4_l2);
        if (has_cols)
          decode_block(pbi, xd, mi_row, mi_col + hbs, r, subsize, n8x8_l2,
                       n4x4_l2);
        break;
      case PARTITION_SPLIT:
        decode_partition(pbi, xd, mi_row, mi_col, r, subsize, n8x8_l2);
        decode_partition(pbi, xd, mi_row, mi_col + hbs, r, subsize, n8x8_l2);
        decode_partition(pbi, xd, mi_row + hbs, mi_col, r, subsize, n8x8_l2);
        decode_partition(pbi, xd, mi_row + hbs, mi_col + hbs, r, subsize,
                         n8x8_l2);
        break;
      default:
        assert(0 && "Invalid partition type");
    }
  }