vp9_rdopt.c 177 KB
Newer Older
John Koleszar's avatar
John Koleszar committed
1
/*
2
 *  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
John Koleszar's avatar
John Koleszar committed
3
 *
4
 *  Use of this source code is governed by a BSD-style license
5
6
 *  that can be found in the LICENSE file in the root of the source
 *  tree. An additional intellectual property rights grant can be found
7
 *  in the file PATENTS.  All contributing project authors may
8
 *  be found in the AUTHORS file in the root of the source tree.
John Koleszar's avatar
John Koleszar committed
9
10
11
12
13
14
15
 */


#include <stdio.h>
#include <math.h>
#include <limits.h>
#include <assert.h>
16
17
#include "vp9/common/vp9_pragmas.h"

18
19
20
21
22
#include "vp9/encoder/vp9_tokenize.h"
#include "vp9/encoder/vp9_treewriter.h"
#include "vp9/encoder/vp9_onyx_int.h"
#include "vp9/encoder/vp9_modecosts.h"
#include "vp9/encoder/vp9_encodeintra.h"
23
24
25
26
27
#include "vp9/common/vp9_entropymode.h"
#include "vp9/common/vp9_reconinter.h"
#include "vp9/common/vp9_reconintra.h"
#include "vp9/common/vp9_findnearmv.h"
#include "vp9/common/vp9_quant_common.h"
28
29
30
31
32
33
#include "vp9/encoder/vp9_encodemb.h"
#include "vp9/encoder/vp9_quantize.h"
#include "vp9/encoder/vp9_variance.h"
#include "vp9/encoder/vp9_mcomp.h"
#include "vp9/encoder/vp9_rdopt.h"
#include "vp9/encoder/vp9_ratectrl.h"
John Koleszar's avatar
John Koleszar committed
34
#include "vpx_mem/vpx_mem.h"
35
36
#include "vp9/common/vp9_systemdependent.h"
#include "vp9/encoder/vp9_encodemv.h"
John Koleszar's avatar
John Koleszar committed
37

38
39
40
#include "vp9/common/vp9_seg_common.h"
#include "vp9/common/vp9_pred_common.h"
#include "vp9/common/vp9_entropy.h"
41
#include "vp9_rtcd.h"
42
#include "vp9/common/vp9_mvref_common.h"
Ronald S. Bultje's avatar
Ronald S. Bultje committed
43
#include "vp9/common/vp9_common.h"
Paul Wilkins's avatar
Paul Wilkins committed
44

John Koleszar's avatar
John Koleszar committed
45
46
#define MAXF(a,b)            (((a) > (b)) ? (a) : (b))

47
48
#define INVALID_MV 0x80008000

49
50
51
/* Factor to weigh the rate for switchable interp filters */
#define SWITCHABLE_INTERP_RATE_FACTOR 1

John Koleszar's avatar
John Koleszar committed
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
static const int auto_speed_thresh[17] = {
  1000,
  200,
  150,
  130,
  150,
  125,
  120,
  115,
  115,
  115,
  115,
  115,
  115,
  115,
  115,
  115,
  105
John Koleszar's avatar
John Koleszar committed
70
71
};

72
const MODE_DEFINITION vp9_mode_order[MAX_MODES] = {
73
74
  {ZEROMV,    LAST_FRAME,   NONE},
  {DC_PRED,   INTRA_FRAME,  NONE},
John Koleszar's avatar
John Koleszar committed
75

76
77
  {NEARESTMV, LAST_FRAME,   NONE},
  {NEARMV,    LAST_FRAME,   NONE},
John Koleszar's avatar
John Koleszar committed
78

79
80
  {ZEROMV,    GOLDEN_FRAME, NONE},
  {NEARESTMV, GOLDEN_FRAME, NONE},
John Koleszar's avatar
John Koleszar committed
81

82
83
  {ZEROMV,    ALTREF_FRAME, NONE},
  {NEARESTMV, ALTREF_FRAME, NONE},
John Koleszar's avatar
John Koleszar committed
84

85
86
  {NEARMV,    GOLDEN_FRAME, NONE},
  {NEARMV,    ALTREF_FRAME, NONE},
John Koleszar's avatar
John Koleszar committed
87

88
89
90
91
92
93
94
95
  {V_PRED,    INTRA_FRAME,  NONE},
  {H_PRED,    INTRA_FRAME,  NONE},
  {D45_PRED,  INTRA_FRAME,  NONE},
  {D135_PRED, INTRA_FRAME,  NONE},
  {D117_PRED, INTRA_FRAME,  NONE},
  {D153_PRED, INTRA_FRAME,  NONE},
  {D27_PRED,  INTRA_FRAME,  NONE},
  {D63_PRED,  INTRA_FRAME,  NONE},
John Koleszar's avatar
John Koleszar committed
96

97
  {TM_PRED,   INTRA_FRAME,  NONE},
John Koleszar's avatar
John Koleszar committed
98

99
100
101
  {NEWMV,     LAST_FRAME,   NONE},
  {NEWMV,     GOLDEN_FRAME, NONE},
  {NEWMV,     ALTREF_FRAME, NONE},
John Koleszar's avatar
John Koleszar committed
102

103
104
105
  {SPLITMV,   LAST_FRAME,   NONE},
  {SPLITMV,   GOLDEN_FRAME, NONE},
  {SPLITMV,   ALTREF_FRAME, NONE},
106

Yaowu Xu's avatar
Yaowu Xu committed
107
  {I4X4_PRED,    INTRA_FRAME,  NONE},
108
  {I8X8_PRED, INTRA_FRAME,  NONE},
109

John Koleszar's avatar
John Koleszar committed
110
111
112
113
  /* compound prediction modes */
  {ZEROMV,    LAST_FRAME,   GOLDEN_FRAME},
  {NEARESTMV, LAST_FRAME,   GOLDEN_FRAME},
  {NEARMV,    LAST_FRAME,   GOLDEN_FRAME},
114

John Koleszar's avatar
John Koleszar committed
115
116
117
  {ZEROMV,    ALTREF_FRAME, LAST_FRAME},
  {NEARESTMV, ALTREF_FRAME, LAST_FRAME},
  {NEARMV,    ALTREF_FRAME, LAST_FRAME},
118

John Koleszar's avatar
John Koleszar committed
119
120
121
  {ZEROMV,    GOLDEN_FRAME, ALTREF_FRAME},
  {NEARESTMV, GOLDEN_FRAME, ALTREF_FRAME},
  {NEARMV,    GOLDEN_FRAME, ALTREF_FRAME},
122

John Koleszar's avatar
John Koleszar committed
123
124
125
  {NEWMV,     LAST_FRAME,   GOLDEN_FRAME},
  {NEWMV,     ALTREF_FRAME, LAST_FRAME  },
  {NEWMV,     GOLDEN_FRAME, ALTREF_FRAME},
126

John Koleszar's avatar
John Koleszar committed
127
128
  {SPLITMV,   LAST_FRAME,   GOLDEN_FRAME},
  {SPLITMV,   ALTREF_FRAME, LAST_FRAME  },
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
  {SPLITMV,   GOLDEN_FRAME, ALTREF_FRAME},

#if CONFIG_COMP_INTERINTRA_PRED
  /* compound inter-intra prediction */
  {ZEROMV,    LAST_FRAME,   INTRA_FRAME},
  {NEARESTMV, LAST_FRAME,   INTRA_FRAME},
  {NEARMV,    LAST_FRAME,   INTRA_FRAME},
  {NEWMV,     LAST_FRAME,   INTRA_FRAME},

  {ZEROMV,    GOLDEN_FRAME,   INTRA_FRAME},
  {NEARESTMV, GOLDEN_FRAME,   INTRA_FRAME},
  {NEARMV,    GOLDEN_FRAME,   INTRA_FRAME},
  {NEWMV,     GOLDEN_FRAME,   INTRA_FRAME},

  {ZEROMV,    ALTREF_FRAME,   INTRA_FRAME},
  {NEARESTMV, ALTREF_FRAME,   INTRA_FRAME},
  {NEARMV,    ALTREF_FRAME,   INTRA_FRAME},
  {NEWMV,     ALTREF_FRAME,   INTRA_FRAME},
#endif
John Koleszar's avatar
John Koleszar committed
148
149
};

150
151
static void fill_token_costs(vp9_coeff_count *c,
                             vp9_coeff_probs *p,
152
                             TX_SIZE tx_size) {
153
  int i, j, k, l;
John Koleszar's avatar
John Koleszar committed
154

155
  for (i = 0; i < BLOCK_TYPES; i++)
156
157
158
    for (j = 0; j < REF_TYPES; j++)
      for (k = 0; k < COEF_BANDS; k++)
        for (l = 0; l < PREV_COEF_CONTEXTS; l++) {
159
160
161
162
163
164
          vp9_cost_tokens_skip((int *)(c[i][j][k][l]),
                               p[i][j][k][l],
                               vp9_coef_tree);
        }
}

165
166
167
168
static int rd_iifactor[32] =  { 4, 4, 3, 2, 1, 0, 0, 0,
                                0, 0, 0, 0, 0, 0, 0, 0,
                                0, 0, 0, 0, 0, 0, 0, 0,
                                0, 0, 0, 0, 0, 0, 0, 0, };
John Koleszar's avatar
John Koleszar committed
169

170
// 3* dc_qlookup[Q]*dc_qlookup[Q];
171

172
/* values are now correlated to quantizer */
Paul Wilkins's avatar
Paul Wilkins committed
173
174
175
static int sad_per_bit16lut[QINDEX_RANGE];
static int sad_per_bit4lut[QINDEX_RANGE];

176
void vp9_init_me_luts() {
John Koleszar's avatar
John Koleszar committed
177
178
179
180
181
182
183
  int i;

  // Initialize the sad lut tables using a formulaic calculation for now
  // This is to make it easier to resolve the impact of experimental changes
  // to the quantizer tables.
  for (i = 0; i < QINDEX_RANGE; i++) {
    sad_per_bit16lut[i] =
184
185
      (int)((0.0418 * vp9_convert_qindex_to_q(i)) + 2.4107);
    sad_per_bit4lut[i] = (int)((0.063 * vp9_convert_qindex_to_q(i)) + 2.742);
John Koleszar's avatar
John Koleszar committed
186
  }
Paul Wilkins's avatar
Paul Wilkins committed
187
}
John Koleszar's avatar
John Koleszar committed
188

189
static int compute_rd_mult(int qindex) {
Dmitry Kovalev's avatar
Dmitry Kovalev committed
190
  int q = vp9_dc_quant(qindex, 0);
191
  return (11 * q * q) >> 2;
192
193
}

Dmitry Kovalev's avatar
Dmitry Kovalev committed
194
195
196
void vp9_initialize_me_consts(VP9_COMP *cpi, int qindex) {
  cpi->mb.sadperbit16 = sad_per_bit16lut[qindex];
  cpi->mb.sadperbit4 = sad_per_bit4lut[qindex];
John Koleszar's avatar
John Koleszar committed
197
198
}

199

Dmitry Kovalev's avatar
Dmitry Kovalev committed
200
void vp9_initialize_rd_consts(VP9_COMP *cpi, int qindex) {
201
  int q, i;
John Koleszar's avatar
John Koleszar committed
202

203
  vp9_clear_system_state();  // __asm emms;
John Koleszar's avatar
John Koleszar committed
204

John Koleszar's avatar
John Koleszar committed
205
206
207
208
  // Further tests required to see if optimum is different
  // for key frames, golden frames and arf frames.
  // if (cpi->common.refresh_golden_frame ||
  //     cpi->common.refresh_alt_ref_frame)
Dmitry Kovalev's avatar
Dmitry Kovalev committed
209
  qindex = (qindex < 0) ? 0 : ((qindex > MAXQ) ? MAXQ : qindex);
210

Dmitry Kovalev's avatar
Dmitry Kovalev committed
211
  cpi->RDMULT = compute_rd_mult(qindex);
John Koleszar's avatar
John Koleszar committed
212
213
214
215
216
  if (cpi->pass == 2 && (cpi->common.frame_type != KEY_FRAME)) {
    if (cpi->twopass.next_iiratio > 31)
      cpi->RDMULT += (cpi->RDMULT * rd_iifactor[31]) >> 4;
    else
      cpi->RDMULT +=
Dmitry Kovalev's avatar
Dmitry Kovalev committed
217
          (cpi->RDMULT * rd_iifactor[cpi->twopass.next_iiratio]) >> 4;
John Koleszar's avatar
John Koleszar committed
218
  }
219
  cpi->mb.errorperbit = cpi->RDMULT >> 6;
John Koleszar's avatar
John Koleszar committed
220
  cpi->mb.errorperbit += (cpi->mb.errorperbit == 0);
221

222
  vp9_set_speed_features(cpi);
John Koleszar's avatar
John Koleszar committed
223

Dmitry Kovalev's avatar
Dmitry Kovalev committed
224
225
  q = (int)pow(vp9_dc_quant(qindex, 0) >> 2, 1.25);
  q <<= 2;
John Koleszar's avatar
John Koleszar committed
226
227
  if (q < 8)
    q = 8;
228

John Koleszar's avatar
John Koleszar committed
229
230
231
  if (cpi->RDMULT > 1000) {
    cpi->RDDIV = 1;
    cpi->RDMULT /= 100;
John Koleszar's avatar
John Koleszar committed
232

John Koleszar's avatar
John Koleszar committed
233
234
235
236
237
238
239
    for (i = 0; i < MAX_MODES; i++) {
      if (cpi->sf.thresh_mult[i] < INT_MAX) {
        cpi->rd_threshes[i] = cpi->sf.thresh_mult[i] * q / 100;
      } else {
        cpi->rd_threshes[i] = INT_MAX;
      }
      cpi->rd_baseline_thresh[i] = cpi->rd_threshes[i];
John Koleszar's avatar
John Koleszar committed
240
    }
John Koleszar's avatar
John Koleszar committed
241
242
  } else {
    cpi->RDDIV = 100;
John Koleszar's avatar
John Koleszar committed
243

John Koleszar's avatar
John Koleszar committed
244
245
246
247
248
249
250
    for (i = 0; i < MAX_MODES; i++) {
      if (cpi->sf.thresh_mult[i] < (INT_MAX / q)) {
        cpi->rd_threshes[i] = cpi->sf.thresh_mult[i] * q;
      } else {
        cpi->rd_threshes[i] = INT_MAX;
      }
      cpi->rd_baseline_thresh[i] = cpi->rd_threshes[i];
John Koleszar's avatar
John Koleszar committed
251
    }
John Koleszar's avatar
John Koleszar committed
252
  }
John Koleszar's avatar
John Koleszar committed
253

254
  fill_token_costs(cpi->mb.token_costs[TX_4X4],
255
                   cpi->common.fc.coef_probs_4x4, TX_4X4);
256
  fill_token_costs(cpi->mb.token_costs[TX_8X8],
257
                   cpi->common.fc.coef_probs_8x8, TX_8X8);
258
  fill_token_costs(cpi->mb.token_costs[TX_16X16],
259
                   cpi->common.fc.coef_probs_16x16, TX_16X16);
260
  fill_token_costs(cpi->mb.token_costs[TX_32X32],
261
                   cpi->common.fc.coef_probs_32x32, TX_32X32);
262

263
  for (i = 0; i < NUM_PARTITION_CONTEXTS; i++)
264
265
266
267
    vp9_cost_tokens(cpi->mb.partition_cost[i],
                    cpi->common.fc.partition_prob[i],
                    vp9_partition_tree);

John Koleszar's avatar
John Koleszar committed
268
269
  /*rough estimate for costing*/
  cpi->common.kf_ymode_probs_index = cpi->common.base_qindex >> 4;
270
  vp9_init_mode_costs(cpi);
John Koleszar's avatar
John Koleszar committed
271

272
  if (cpi->common.frame_type != KEY_FRAME) {
273
    vp9_build_nmv_cost_table(
274
275
276
277
278
279
        cpi->mb.nmvjointcost,
        cpi->mb.e_mbd.allow_high_precision_mv ?
        cpi->mb.nmvcost_hp : cpi->mb.nmvcost,
        &cpi->common.fc.nmvc,
        cpi->mb.e_mbd.allow_high_precision_mv, 1, 1);
  }
John Koleszar's avatar
John Koleszar committed
280
281
}

282
int vp9_block_error_c(int16_t *coeff, int16_t *dqcoeff, int block_size) {
283
  int i, error = 0;
John Koleszar's avatar
John Koleszar committed
284

285
  for (i = 0; i < block_size; i++) {
John Koleszar's avatar
John Koleszar committed
286
287
288
    int this_diff = coeff[i] - dqcoeff[i];
    error += this_diff * this_diff;
  }
John Koleszar's avatar
John Koleszar committed
289

John Koleszar's avatar
John Koleszar committed
290
  return error;
John Koleszar's avatar
John Koleszar committed
291
292
}

293
static INLINE int cost_coeffs(VP9_COMMON *const cm, MACROBLOCK *mb,
294
                              int ib, PLANE_TYPE type,
295
296
                              ENTROPY_CONTEXT *a,
                              ENTROPY_CONTEXT *l,
John Koleszar's avatar
John Koleszar committed
297
298
                              TX_SIZE tx_size,
                              int y_blocks) {
299
  MACROBLOCKD *const xd = &mb->e_mbd;
300
301
  MB_MODE_INFO *mbmi = &xd->mode_info_context->mbmi;
  int pt;
302
  int c = 0;
303
304
  int cost = 0, pad;
  const int *scan, *nb;
John Koleszar's avatar
John Koleszar committed
305
306
  const struct plane_block_idx pb_idx = plane_block_idx(y_blocks, ib);
  const int eob = xd->plane[pb_idx.plane].eobs[pb_idx.block];
307
308
  const int16_t *qcoeff_ptr = BLOCK_OFFSET(xd->plane[pb_idx.plane].qcoeff,
                                           pb_idx.block, 16);
309
  const int ref = mbmi->ref_frame != INTRA_FRAME;
310
  unsigned int (*token_costs)[PREV_COEF_CONTEXTS][MAX_ENTROPY_TOKENS] =
311
      mb->token_costs[tx_size][type][ref];
312
  ENTROPY_CONTEXT a_ec, l_ec;
313
314
315
316
  ENTROPY_CONTEXT *const a1 = a +
      sizeof(ENTROPY_CONTEXT_PLANES)/sizeof(ENTROPY_CONTEXT);
  ENTROPY_CONTEXT *const l1 = l +
      sizeof(ENTROPY_CONTEXT_PLANES)/sizeof(ENTROPY_CONTEXT);
317
  TX_TYPE tx_type = DCT_DCT;
318

319
320
321
322
323
324
325
326
#if CONFIG_CODE_ZEROGROUP
  int last_nz_pos[3] = {-1, -1, -1};  // Encoder only
  int is_eoo_list[3] = {0, 0, 0};
  int is_eoo_negative[3] = {0, 0, 0};
  int is_last_zero[3] = {0, 0, 0};
  int o, rc, skip_coef_val;
  vp9_zpc_probs *zpc_probs;
  uint8_t token_cache_full[1024];
327
#endif
328
  const int segment_id = xd->mode_info_context->mbmi.segment_id;
329
330
331
332
  vp9_prob (*coef_probs)[REF_TYPES][COEF_BANDS][PREV_COEF_CONTEXTS]
                        [ENTROPY_NODES];
  int seg_eob, default_eob;
  uint8_t token_cache[1024];
333

334
335
336
337
#if CONFIG_CODE_ZEROGROUP
  vpx_memset(token_cache, UNKNOWN_TOKEN, sizeof(token_cache));
#endif

338
  // Check for consistency of tx_size with mode info
John Koleszar's avatar
John Koleszar committed
339
  assert((!type && !pb_idx.plane) || (type && pb_idx.plane));
340
341
342
343
344
345
346
  if (type == PLANE_TYPE_Y_WITH_DC) {
    assert(xd->mode_info_context->mbmi.txfm_size == tx_size);
  } else {
    TX_SIZE tx_size_uv = get_uv_tx_size(xd);
    assert(tx_size == tx_size_uv);
  }

347
  switch (tx_size) {
348
    case TX_4X4: {
349
350
      tx_type = (type == PLANE_TYPE_Y_WITH_DC) ?
          get_tx_type_4x4(xd, ib) : DCT_DCT;
351
352
      a_ec = *a;
      l_ec = *l;
353
      coef_probs = cm->fc.coef_probs_4x4;
354
      seg_eob = 16;
355
      scan = get_scan_4x4(tx_type);
356
357
358
#if CONFIG_CODE_ZEROGROUP
      zpc_probs = &cm->fc.zpc_probs_4x4;
#endif
Daniel Kang's avatar
Daniel Kang committed
359
      break;
360
    }
361
362
    case TX_8X8: {
      const BLOCK_SIZE_TYPE sb_type = xd->mode_info_context->mbmi.sb_type;
363
      const int sz = 1 + b_width_log2(sb_type);
364
      const int x = ib & ((1 << sz) - 1), y = ib - x;
365
366
      TX_TYPE tx_type = (type == PLANE_TYPE_Y_WITH_DC) ?
          get_tx_type_8x8(xd, y + (x >> 1)) : DCT_DCT;
367
368
      a_ec = (a[0] + a[1]) != 0;
      l_ec = (l[0] + l[1]) != 0;
369
      scan = get_scan_8x8(tx_type);
370
      coef_probs = cm->fc.coef_probs_8x8;
371
      seg_eob = 64;
372
373
374
#if CONFIG_CODE_ZEROGROUP
      zpc_probs = &cm->fc.zpc_probs_8x8;
#endif
Daniel Kang's avatar
Daniel Kang committed
375
      break;
376
377
378
    }
    case TX_16X16: {
      const BLOCK_SIZE_TYPE sb_type = xd->mode_info_context->mbmi.sb_type;
379
      const int sz = 2 + b_width_log2(sb_type);
380
      const int x = ib & ((1 << sz) - 1), y = ib - x;
381
382
      TX_TYPE tx_type = (type == PLANE_TYPE_Y_WITH_DC) ?
          get_tx_type_16x16(xd, y + (x >> 2)) : DCT_DCT;
383
      scan = get_scan_16x16(tx_type);
384
      coef_probs = cm->fc.coef_probs_16x16;
385
      seg_eob = 256;
386
      if (type == PLANE_TYPE_UV) {
387
388
389
390
391
        a_ec = (a[0] + a[1] + a1[0] + a1[1]) != 0;
        l_ec = (l[0] + l[1] + l1[0] + l1[1]) != 0;
      } else {
        a_ec = (a[0] + a[1] + a[2] + a[3]) != 0;
        l_ec = (l[0] + l[1] + l[2] + l[3]) != 0;
Deb Mukherjee's avatar
Deb Mukherjee committed
392
      }
393
394
395
#if CONFIG_CODE_ZEROGROUP
      zpc_probs = &cm->fc.zpc_probs_16x16;
#endif
Daniel Kang's avatar
Daniel Kang committed
396
      break;
397
    }
398
399
    case TX_32X32:
      scan = vp9_default_zig_zag1d_32x32;
400
      coef_probs = cm->fc.coef_probs_32x32;
401
      seg_eob = 1024;
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
      if (type == PLANE_TYPE_UV) {
        ENTROPY_CONTEXT *a2, *a3, *l2, *l3;
        a2 = a1 + sizeof(ENTROPY_CONTEXT_PLANES) / sizeof(ENTROPY_CONTEXT);
        a3 = a2 + sizeof(ENTROPY_CONTEXT_PLANES) / sizeof(ENTROPY_CONTEXT);
        l2 = l1 + sizeof(ENTROPY_CONTEXT_PLANES) / sizeof(ENTROPY_CONTEXT);
        l3 = l2 + sizeof(ENTROPY_CONTEXT_PLANES) / sizeof(ENTROPY_CONTEXT);
        a_ec = (a[0] + a[1] + a1[0] + a1[1] +
                a2[0] + a2[1] + a3[0] + a3[1]) != 0;
        l_ec = (l[0] + l[1] + l1[0] + l1[1] +
                l2[0] + l2[1] + l3[0] + l3[1]) != 0;
      } else {
        a_ec = (a[0] + a[1] + a[2] + a[3] +
                a1[0] + a1[1] + a1[2] + a1[3]) != 0;
        l_ec = (l[0] + l[1] + l[2] + l[3] +
                l1[0] + l1[1] + l1[2] + l1[3]) != 0;
      }
418
419
420
#if CONFIG_CODE_ZEROGROUP
      zpc_probs = &cm->fc.zpc_probs_32x32;
#endif
421
      break;
Daniel Kang's avatar
Daniel Kang committed
422
    default:
423
      abort();
Daniel Kang's avatar
Daniel Kang committed
424
425
      break;
  }
John Koleszar's avatar
John Koleszar committed
426
  assert(eob <= seg_eob);
427

428
  pt = combine_entropy_contexts(a_ec, l_ec);
429
430
  nb = vp9_get_coef_neighbors_handle(scan, &pad);
  default_eob = seg_eob;
431

432
433
  if (vp9_segfeature_active(xd, segment_id, SEG_LVL_SKIP))
    seg_eob = 0;
434

435
436
437
438
  /* sanity check to ensure that we do not have spurious non-zero q values */
  if (eob < seg_eob)
    assert(qcoeff_ptr[scan[eob]] == 0);

439
440
441
442
443
444
445
446
447
448
#if CONFIG_CODE_ZEROGROUP
  vpx_memset(token_cache_full, ZERO_TOKEN, sizeof(token_cache_full));
  for (c = 0; c < eob; ++c) {
    rc = scan[c];
    token_cache_full[rc] = vp9_dct_value_tokens_ptr[qcoeff_ptr[rc]].token;
    o = vp9_get_orientation(rc, tx_size);
    if (qcoeff_ptr[rc] != 0)
      last_nz_pos[o] = c;
  }
#endif
449
  {
450
    for (c = 0; c < eob; c++) {
451
      int v = qcoeff_ptr[scan[c]];
452
      int t = vp9_dct_value_tokens_ptr[v].token;
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
      int band = get_coef_band(scan, tx_size, c);
      if (c)
        pt = vp9_get_coef_context(scan, nb, pad, token_cache, c, default_eob);
#if CONFIG_CODE_ZEROGROUP
      rc = scan[c];
      o = vp9_get_orientation(rc, tx_size);
      skip_coef_val = (token_cache[rc] == ZERO_TOKEN || is_eoo_list[o]);
      if (!skip_coef_val) {
        cost += token_costs[band][pt][t] + vp9_dct_value_cost_ptr[v];
      } else {
        assert(v == 0);
      }
#else
      cost += token_costs[band][pt][t] + vp9_dct_value_cost_ptr[v];
#endif
468
469
      if (!c || token_cache[scan[c - 1]])
        cost += vp9_cost_bit(coef_probs[type][ref][band][pt][0], 1);
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
      token_cache[scan[c]] = t;
#if CONFIG_CODE_ZEROGROUP
      if (t == ZERO_TOKEN && !skip_coef_val) {
        int eoo = 0, use_eoo;
#if USE_ZPC_EOORIENT == 1
        use_eoo = vp9_use_eoo(c, seg_eob, scan, tx_size,
                              is_last_zero, is_eoo_list);
#else
        use_eoo = 0;
#endif
        if (use_eoo) {
          eoo = vp9_is_eoo(c, eob, scan, tx_size, qcoeff_ptr, last_nz_pos);
          if (eoo && is_eoo_negative[o]) eoo = 0;
          if (eoo) {
            int c_;
            int savings = 0;
            int zsaved = 0;
            savings = vp9_cost_bit((*zpc_probs)[ref]
                                   [coef_to_zpc_band(band)]
                                   [coef_to_zpc_ptok(pt)][0], 1) -
                      vp9_cost_bit((*zpc_probs)[ref]
                                   [coef_to_zpc_band(band)]
                                   [coef_to_zpc_ptok(pt)][0], 0);
            for (c_ = c + 1; c_ < eob; ++c_) {
              if (o == vp9_get_orientation(scan[c_], tx_size)) {
                int pt_ = vp9_get_coef_context(scan, nb, pad,
                                               token_cache_full, c_,
                                               default_eob);
                int band_ = get_coef_band(scan, tx_size, c_);
                assert(token_cache_full[scan[c_]] == ZERO_TOKEN);
                if (!c_ || token_cache_full[scan[c_ - 1]])
                  savings += vp9_cost_bit(
                      coef_probs[type][ref][band_][pt_][0], 1);
                savings += vp9_cost_bit(
                    coef_probs[type][ref][band_][pt_][1], 0);
                zsaved++;
              }
            }
            if (savings < 0) {
            // if (zsaved < ZPC_ZEROSSAVED_EOO) {
              eoo = 0;
              is_eoo_negative[o] = 1;
            }
          }
        }
        if (use_eoo) {
          cost += vp9_cost_bit((*zpc_probs)[ref]
                                           [coef_to_zpc_band(band)]
                                           [coef_to_zpc_ptok(pt)][0], !eoo);
          if (eoo) {
            assert(is_eoo_list[o] == 0);
            is_eoo_list[o] = 1;
          }
        }
      }
      is_last_zero[o] = (t == ZERO_TOKEN);
526
#endif
527
    }
528
529
530
531
532
533
534
    if (c < seg_eob) {
      if (c)
        pt = vp9_get_coef_context(scan, nb, pad, token_cache, c, default_eob);
      cost += mb->token_costs[tx_size][type][ref]
          [get_coef_band(scan, tx_size, c)]
          [pt][DCT_EOB_TOKEN];
    }
535
536
  }

537
538
    // is eob first coefficient;
    pt = (c > 0);
539
  *a = *l = pt;
540
541
542
543
544
545
546
547
548
549
550
551
552
553
  if (tx_size >= TX_8X8) {
    a[1] = l[1] = pt;
    if (tx_size >= TX_16X16) {
      if (type == PLANE_TYPE_UV) {
        a1[0] = a1[1] = l1[0] = l1[1] = pt;
      } else {
        a[2] = a[3] = l[2] = l[3] = pt;
        if (tx_size >= TX_32X32) {
          a1[0] = a1[1] = a1[2] = a1[3] = pt;
          l1[0] = l1[1] = l1[2] = l1[3] = pt;
        }
      }
    }
  }
554
555
556
  return cost;
}

557
static void choose_txfm_size_from_rd(VP9_COMP *cpi, MACROBLOCK *x,
558
559
560
561
562
                                     int (*r)[2], int *rate,
                                     int *d, int *distortion,
                                     int *s, int *skip,
                                     int64_t txfm_cache[NB_TXFM_MODES],
                                     TX_SIZE max_txfm_size) {
563
564
565
  VP9_COMMON *const cm = &cpi->common;
  MACROBLOCKD *const xd = &x->e_mbd;
  MB_MODE_INFO *const mbmi = &xd->mode_info_context->mbmi;
566
  vp9_prob skip_prob = vp9_get_pred_prob(cm, xd, PRED_MBSKIP);
567
568
  int64_t rd[TX_SIZE_MAX_SB][2];
  int n, m;
569
  int s0, s1;
570
571
572
573
574
575
576
577
578
579

  for (n = TX_4X4; n <= max_txfm_size; n++) {
    r[n][1] = r[n][0];
    for (m = 0; m <= n - (n == max_txfm_size); m++) {
      if (m == n)
        r[n][1] += vp9_cost_zero(cm->prob_tx[m]);
      else
        r[n][1] += vp9_cost_one(cm->prob_tx[m]);
    }
  }
580

581
582
583
  assert(skip_prob > 0);
  s0 = vp9_cost_bit(skip_prob, 0);
  s1 = vp9_cost_bit(skip_prob, 1);
584

585
586
587
588
589
590
  for (n = TX_4X4; n <= max_txfm_size; n++) {
    if (s[n]) {
      rd[n][0] = rd[n][1] = RDCOST(x->rdmult, x->rddiv, s1, d[n]);
    } else {
      rd[n][0] = RDCOST(x->rdmult, x->rddiv, r[n][0] + s0, d[n]);
      rd[n][1] = RDCOST(x->rdmult, x->rddiv, r[n][1] + s0, d[n]);
591
592
593
    }
  }

594
595
596
597
598
599
  if (max_txfm_size == TX_32X32 &&
      (cm->txfm_mode == ALLOW_32X32 ||
       (cm->txfm_mode == TX_MODE_SELECT &&
        rd[TX_32X32][1] < rd[TX_16X16][1] && rd[TX_32X32][1] < rd[TX_8X8][1] &&
        rd[TX_32X32][1] < rd[TX_4X4][1]))) {
    mbmi->txfm_size = TX_32X32;
Ronald S. Bultje's avatar
Ronald S. Bultje committed
600
601
602
603
604
  } else if ( cm->txfm_mode == ALLOW_16X16 ||
             (max_txfm_size == TX_16X16 && cm->txfm_mode == ALLOW_32X32) ||
             (cm->txfm_mode == TX_MODE_SELECT &&
              rd[TX_16X16][1] < rd[TX_8X8][1] &&
              rd[TX_16X16][1] < rd[TX_4X4][1])) {
605
    mbmi->txfm_size = TX_16X16;
606
  } else if (cm->txfm_mode == ALLOW_8X8 ||
607
           (cm->txfm_mode == TX_MODE_SELECT && rd[TX_8X8][1] < rd[TX_4X4][1])) {
608
609
    mbmi->txfm_size = TX_8X8;
  } else {
610
    assert(cm->txfm_mode == ONLY_4X4 || cm->txfm_mode == TX_MODE_SELECT);
611
612
613
    mbmi->txfm_size = TX_4X4;
  }

614
  *distortion = d[mbmi->txfm_size];
615
  *rate       = r[mbmi->txfm_size][cm->txfm_mode == TX_MODE_SELECT];
616
617
  *skip       = s[mbmi->txfm_size];

618
619
620
621
622
623
624
625
  txfm_cache[ONLY_4X4] = rd[TX_4X4][0];
  txfm_cache[ALLOW_8X8] = rd[TX_8X8][0];
  txfm_cache[ALLOW_16X16] = rd[TX_16X16][0];
  txfm_cache[ALLOW_32X32] = rd[max_txfm_size][0];
  if (max_txfm_size == TX_32X32 &&
      rd[TX_32X32][1] < rd[TX_16X16][1] && rd[TX_32X32][1] < rd[TX_8X8][1] &&
      rd[TX_32X32][1] < rd[TX_4X4][1])
    txfm_cache[TX_MODE_SELECT] = rd[TX_32X32][1];
Ronald S. Bultje's avatar
Ronald S. Bultje committed
626
  else if (rd[TX_16X16][1] < rd[TX_8X8][1] && rd[TX_16X16][1] < rd[TX_4X4][1])
627
    txfm_cache[TX_MODE_SELECT] = rd[TX_16X16][1];
628
  else
629
630
    txfm_cache[TX_MODE_SELECT] = rd[TX_4X4][1] < rd[TX_8X8][1] ?
                                 rd[TX_4X4][1] : rd[TX_8X8][1];
631
632
}

633
634
static int block_error(int16_t *coeff, int16_t *dqcoeff,
                       int block_size, int shift) {
635
636
637
638
  int i;
  int64_t error = 0;

  for (i = 0; i < block_size; i++) {
639
    int this_diff = coeff[i] - dqcoeff[i];
640
641
    error += this_diff * this_diff;
  }
642
  error >>= shift;
643

Frank Galligan's avatar
Frank Galligan committed
644
  return error > INT_MAX ? INT_MAX : (int)error;
645
646
}

647
648
649
650
static int block_error_sby(MACROBLOCK *x, int block_size, int shift) {
  return block_error(x->plane[0].coeff, x->e_mbd.plane[0].dqcoeff,
                     block_size, shift);
}
651

652
653
654
655
static int block_error_sbuv(MACROBLOCK *x, BLOCK_SIZE_TYPE bsize, int shift) {
  const int bwl = b_width_log2(bsize), bhl = b_height_log2(bsize);
  int64_t sum = 0;
  int plane;
656

657
658
659
660
661
662
663
664
  for (plane = 1; plane < MAX_MB_PLANE; plane++) {
    const int subsampling = x->e_mbd.plane[plane].subsampling_x +
                            x->e_mbd.plane[plane].subsampling_y;
    sum += block_error(x->plane[plane].coeff, x->e_mbd.plane[plane].dqcoeff,
                       16 << (bwl + bhl - subsampling), 0);
  }
  sum >>= shift;
  return sum > INT_MAX ? INT_MAX : (int)sum;
665
666
}

667
668
static int rdcost_sby_4x4(VP9_COMMON *const cm, MACROBLOCK *x,
                          BLOCK_SIZE_TYPE bsize) {
669
670
  const int bwl = b_width_log2(bsize), bw = 1 << bwl;
  const int bh = 1 << b_height_log2(bsize);
671
672
  int cost = 0, b;
  MACROBLOCKD *const xd = &x->e_mbd;
673
  ENTROPY_CONTEXT_PLANES t_above[4], t_left[4];
674

675
676
677
678
  vpx_memcpy(&t_above, xd->above_context,
             (sizeof(ENTROPY_CONTEXT_PLANES) * bw) >> 2);
  vpx_memcpy(&t_left,  xd->left_context,
             (sizeof(ENTROPY_CONTEXT_PLANES) * bh) >> 2);
679

680
681
  for (b = 0; b < bw * bh; b++) {
    const int x_idx = b & (bw - 1), y_idx = b >> bwl;
682
    cost += cost_coeffs(cm, x, b, PLANE_TYPE_Y_WITH_DC,
683
684
685
686
                ((ENTROPY_CONTEXT *) &t_above[x_idx >> 2]) + (x_idx & 3),
                ((ENTROPY_CONTEXT *) &t_left[y_idx >> 2]) + (y_idx & 3),
                TX_4X4, bw * bh);
  }
687
688
689
690
691

  return cost;
}

static void super_block_yrd_4x4(VP9_COMMON *const cm, MACROBLOCK *x,
692
693
                                int *rate, int *distortion, int *skippable,
                                BLOCK_SIZE_TYPE bsize) {
694
  const int bwl = b_width_log2(bsize), bhl = b_height_log2(bsize);
695
696
697
  MACROBLOCKD *const xd = &x->e_mbd;

  xd->mode_info_context->mbmi.txfm_size = TX_4X4;
698
699
  vp9_transform_sby_4x4(x, bsize);
  vp9_quantize_sby_4x4(x, bsize);
700

701
  *distortion = block_error_sby(x, 16 << (bwl + bhl), 2);
702
  *rate       = rdcost_sby_4x4(cm, x, bsize);
703
  *skippable  = vp9_sby_is_skippable(xd, bsize);
704
705
}

706
707
static int rdcost_sby_8x8(VP9_COMMON *const cm, MACROBLOCK *x,
                          BLOCK_SIZE_TYPE bsize) {
708
709
  const int bwl = b_width_log2(bsize) - 1, bw = 1 << bwl;
  const int bh = 1 << (b_height_log2(bsize) - 1);
710
711
  int cost = 0, b;
  MACROBLOCKD *const xd = &x->e_mbd;
712
  ENTROPY_CONTEXT_PLANES t_above[4], t_left[4];
713

714
715
716
717
718
719
720
721
722
723
724
725
  vpx_memcpy(&t_above, xd->above_context,
             (sizeof(ENTROPY_CONTEXT_PLANES) * bw) >> 1);
  vpx_memcpy(&t_left,  xd->left_context,
             (sizeof(ENTROPY_CONTEXT_PLANES) * bh) >> 1);

  for (b = 0; b < bw * bh; b++) {
    const int x_idx = b & (bw - 1), y_idx = b >> bwl;
    cost += cost_coeffs(cm, x, b * 4, PLANE_TYPE_Y_WITH_DC,
                ((ENTROPY_CONTEXT *) &t_above[x_idx >> 1]) + ((x_idx & 1) << 1),
                ((ENTROPY_CONTEXT *) &t_left[y_idx >> 1]) + ((y_idx & 1) << 1),
                TX_8X8, 4 * bw * bh);
  }
726
727
728
729
730

  return cost;
}

static void super_block_yrd_8x8(VP9_COMMON *const cm, MACROBLOCK *x,
731
732
                                int *rate, int *distortion, int *skippable,
                                BLOCK_SIZE_TYPE bsize) {
733
  const int bwl = b_width_log2(bsize), bhl = b_height_log2(bsize);
734
735
736
  MACROBLOCKD *const xd = &x->e_mbd;

  xd->mode_info_context->mbmi.txfm_size = TX_8X8;
737
738
  vp9_transform_sby_8x8(x, bsize);
  vp9_quantize_sby_8x8(x, bsize);
739

740
  *distortion = block_error_sby(x, 16 << (bhl + bwl), 2);
741
  *rate       = rdcost_sby_8x8(cm, x, bsize);
742
  *skippable  = vp9_sby_is_skippable(xd, bsize);
743
744
}

745
746
static int rdcost_sby_16x16(VP9_COMMON *const cm, MACROBLOCK *x,
                            BLOCK_SIZE_TYPE bsize) {
747
748
  const int bwl = b_width_log2(bsize) - 2, bw = 1 << bwl;
  const int bh = 1 << (b_height_log2(bsize) - 2);
749
750
  int cost = 0, b;
  MACROBLOCKD *const xd = &x->e_mbd;
751
  ENTROPY_CONTEXT_PLANES t_above[4], t_left[4];
752

753
754
  vpx_memcpy(&t_above, xd->above_context, sizeof(ENTROPY_CONTEXT_PLANES) * bw);
  vpx_memcpy(&t_left,  xd->left_context,  sizeof(ENTROPY_CONTEXT_PLANES) * bh);
755

756
757
758
759
760
761
762
  for (b = 0; b < bw * bh; b++) {
    const int x_idx = b & (bw - 1), y_idx = b >> bwl;
    cost += cost_coeffs(cm, x, b * 16, PLANE_TYPE_Y_WITH_DC,
                        (ENTROPY_CONTEXT *) &t_above[x_idx],
                        (ENTROPY_CONTEXT *) &t_left[y_idx],
                        TX_16X16, bw * bh * 16);
  }
763
764
765
766
767

  return cost;
}

static void super_block_yrd_16x16(VP9_COMMON *const cm, MACROBLOCK *x,
768
769
                                  int *rate, int *distortion, int *skippable,
                                  BLOCK_SIZE_TYPE bsize) {
770
  const int bwl = b_width_log2(bsize), bhl = b_height_log2(bsize);
771
772
773
  MACROBLOCKD *const xd = &x->e_mbd;

  xd->mode_info_context->mbmi.txfm_size = TX_16X16;
774
775
  vp9_transform_sby_16x16(x, bsize);
  vp9_quantize_sby_16x16(x, bsize);
776

777
  *distortion = block_error_sby(x, 16 << (bwl + bhl), 2);
778
  *rate       = rdcost_sby_16x16(cm, x, bsize);
779
  *skippable  = vp9_sby_is_skippable(xd, bsize);
780
781
}

782
783
static int rdcost_sby_32x32(VP9_COMMON *const cm, MACROBLOCK *x,
                            BLOCK_SIZE_TYPE bsize) {
784
785
  const int bwl = b_width_log2(bsize) - 3, bw = 1 << bwl;
  const int bh = 1 << (b_height_log2(bsize) - 3);
786
  int cost = 0, b;
787
  MACROBLOCKD * const xd = &x->e_mbd;
788
  ENTROPY_CONTEXT_PLANES t_above[4], t_left[4];
789

790
791
792
793
794
795
796
797
798
799
800
801
  vpx_memcpy(&t_above, xd->above_context,
             sizeof(ENTROPY_CONTEXT_PLANES) * bw * 2);
  vpx_memcpy(&t_left,  xd->left_context,
             sizeof(ENTROPY_CONTEXT_PLANES) * bh * 2);

  for (b = 0; b < bw * bh; b++) {
    const int x_idx = b & (bw - 1), y_idx = b >> bwl;
    cost += cost_coeffs(cm, x, b * 64, PLANE_TYPE_Y_WITH_DC,
                        (ENTROPY_CONTEXT *) &t_above[x_idx * 2],
                        (ENTROPY_CONTEXT *) &t_left[y_idx * 2],
                        TX_32X32, bw * bh * 64);
  }
802

803
  return cost;
804
805
}

806
static void super_block_yrd_32x32(VP9_COMMON *const cm, MACROBLOCK *x,
807
808
                                  int *rate, int *distortion, int *skippable,
                                  BLOCK_SIZE_TYPE bsize) {
809
  const int bwl = b_width_log2(bsize), bhl = b_height_log2(bsize);
810
  MACROBLOCKD *const xd = &x->e_mbd;
811

812
  xd->mode_info_context->mbmi.txfm_size = TX_32X32;
813
814
  vp9_transform_sby_32x32(x, bsize);
  vp9_quantize_sby_32x32(x, bsize);
815

816
  *distortion = block_error_sby(x, 16 << (bwl + bhl), 0);
817
  *rate       = rdcost_sby_32x32(cm, x, bsize);
818
  *skippable  = vp9_sby_is_skippable(xd, bsize);
819
820
}

821
822
static void super_block_yrd(VP9_COMP *cpi,
                            MACROBLOCK *x, int *rate, int *distortion,
823
                            int *skip, BLOCK_SIZE_TYPE bs,
824
                            int64_t txfm_cache[NB_TXFM_MODES]) {
825
826
  VP9_COMMON *const cm = &cpi->common;
  int r[TX_SIZE_MAX_SB][2], d[TX_SIZE_MAX_SB], s[TX_SIZE_MAX_SB];
Ronald S. Bultje's avatar
Ronald S. Bultje committed
827

828
  vp9_subtract_sby(x, bs);
Ronald S. Bultje's avatar
Ronald S. Bultje committed
829

830
831
832
833
834
835
  if (bs >= BLOCK_SIZE_SB32X32)
    super_block_yrd_32x32(cm, x, &r[TX_32X32][0], &d[TX_32X32], &s[TX_32X32],
                          bs);
  super_block_yrd_16x16(cm, x, &r[TX_16X16][0], &d[TX_16X16], &s[TX_16X16], bs);
  super_block_yrd_8x8(cm, x,   &r[TX_8X8][0],   &d[TX_8X8],   &s[TX_8X8],   bs);
  super_block_yrd_4x4(cm, x,   &r[TX_4X4][0],   &d[TX_4X4],   &s[TX_4X4],   bs);
Ronald S. Bultje's avatar
Ronald S. Bultje committed
836
837

  choose_txfm_size_from_rd(cpi, x, r, rate, d, distortion, s, skip, txfm_cache,
838
                           TX_32X32 - (bs < BLOCK_SIZE_SB32X32));
Ronald S. Bultje's avatar
Ronald S. Bultje committed
839
}
Ronald S. Bultje's avatar
Ronald S. Bultje committed
840

841
842
static int64_t rd_pick_intra4x4block(VP9_COMP *cpi, MACROBLOCK *x, int ib,
                                     B_PREDICTION_MODE *best_mode,
843
844
845
846
                                     int *bmode_costs,
                                     ENTROPY_CONTEXT *a, ENTROPY_CONTEXT *l,
                                     int *bestrate, int *bestratey,
                                     int *bestdistortion) {
John Koleszar's avatar
John Koleszar committed
847
  B_PREDICTION_MODE mode;
Deb Mukherjee's avatar
Deb Mukherjee committed
848
  MACROBLOCKD *xd = &x->e_mbd;
849
  int64_t best_rd = INT64_MAX;
John Koleszar's avatar
John Koleszar committed
850
851
  int rate = 0;
  int distortion;
852
  VP9_COMMON *const cm = &cpi->common;
John Koleszar's avatar
John Koleszar committed
853
854
855
856
  const int src_stride = x->plane[0].src.stride;
  uint8_t* const src =
      raster_block_offset_uint8(xd, BLOCK_SIZE_MB16X16, 0, ib,
                                x->plane[0].src.buf, src_stride);
857
858
859
  int16_t* const src_diff =
      raster_block_offset_int16(xd, BLOCK_SIZE_MB16X16, 0, ib,
                                x->plane[0].src_diff);
John Koleszar's avatar
John Koleszar committed
860
861
862
  int16_t* const diff =
      raster_block_offset_int16(xd, BLOCK_SIZE_MB16X16, 0, ib,
                                xd->plane[0].diff);
John Koleszar's avatar
John Koleszar committed
863
  int16_t* const coeff = BLOCK_OFFSET(x->plane[0].coeff, ib, 16);
864
865
866
  uint8_t* const dst =
      raster_block_offset_uint8(xd, BLOCK_SIZE_MB16X16, 0, ib,
                                xd->plane[0].dst.buf, xd->plane[0].dst.stride);
John Koleszar's avatar
John Koleszar committed
867
868
  ENTROPY_CONTEXT ta = *a, tempa = *a;
  ENTROPY_CONTEXT tl = *l, templ = *l;
Deb Mukherjee's avatar
Deb Mukherjee committed
869
870
  TX_TYPE tx_type = DCT_DCT;
  TX_TYPE best_tx_type = DCT_DCT;
John Koleszar's avatar
John Koleszar committed
871
872
873
874
875
  /*
   * The predictor buffer is a 2d buffer with a stride of 16.  Create
   * a temp buffer that meets the stride requirements, but we are only
   * interested in the left 4x4 block
   * */
876
  DECLARE_ALIGNED_ARRAY(16, int16_t, best_dqcoeff, 16);
John Koleszar's avatar
John Koleszar committed
877

878
  assert(ib < 16);
879
#if CONFIG_NEWBINTRAMODES
Scott LaVarnway's avatar
Scott LaVarnway committed
880
  xd->mode_info_context->bmi[ib].as_mode.context =
881
    vp9_find_bpred_context(xd, ib, dst, xd->plane[0].dst.stride);
882
#endif
883
  xd->mode_info_context->mbmi.txfm_size = TX_4X4;
884
  for (mode = B_DC_PRED; mode < LEFT4X4; mode++) {
885
886
    int64_t this_rd;
    int ratey;
887
888

#if CONFIG_NEWBINTRAMODES
889
890
891
892
893
894
895
    if (xd->frame_type == KEY_FRAME) {
      if (mode == B_CONTEXT_PRED) continue;
    } else {
      if (mode >= B_CONTEXT_PRED - CONTEXT_PRED_REPLACEMENTS &&
          mode < B_CONTEXT_PRED)
        continue;
    }
896
#endif
897

Scott LaVarnway's avatar
Scott LaVarnway committed
898
    xd->mode_info_context->bmi[ib].as_mode.first = mode;
899
#if CONFIG_NEWBINTRAMODES
900
901
    rate = bmode_costs[
        mode == B_CONTEXT_PRED ? mode - CONTEXT_PRED_REPLACEMENTS : mode];
902
#else
903
    rate = bmode_costs[mode];
904
#endif
John Koleszar's avatar
John Koleszar committed
905

John Koleszar's avatar
John Koleszar committed
906
    vp9_intra4x4_predict(xd, ib, mode, dst, xd->plane[0].dst.stride);
907
    vp9_subtract_block(4, 4, src_diff, 16,
John Koleszar's avatar
John Koleszar committed
908
                       src, src_stride,
909
                       dst, xd->plane[0].dst.stride);
John Koleszar's avatar
John Koleszar committed
910

Scott LaVarnway's avatar
Scott LaVarnway committed
911
    xd->mode_info_context->bmi[ib].as_mode.first = mode;
John Koleszar's avatar
John Koleszar committed
912
    tx_type = get_tx_type_4x4(xd, ib);
913
    if (tx_type != DCT_DCT) {
John Koleszar's avatar
John Koleszar committed
914
      vp9_short_fht4x4(src_diff, coeff, 16, tx_type);
915
      x->quantize_b_4x4(x, ib, tx_type, 16);
916
    } else {
John Koleszar's avatar
John Koleszar committed
917
      x->fwd_txm4x4(src_diff, coeff, 32);
918
      x->quantize_b_4x4(x, ib, tx_type, 16);
919
    }
John Koleszar's avatar
John Koleszar committed
920

921
922
    tempa = ta;
    templ = tl;
Jingning Han's avatar
Jingning Han committed
923

John Koleszar's avatar
John Koleszar committed
924
    ratey = cost_coeffs(cm, x, ib,
John Koleszar's avatar
John Koleszar committed
925
                        PLANE_TYPE_Y_WITH_DC, &tempa, &templ, TX_4X4, 16);
926
    rate += ratey;
John Koleszar's avatar
John Koleszar committed
927
    distortion = vp9_block_error(coeff,
928
929
                                 BLOCK_OFFSET(xd->plane[0].dqcoeff, ib, 16),
                                 16) >> 2;
Jingning Han's avatar
Jingning Han committed
930

931
    this_rd = RDCOST(x->rdmult, x->rddiv, rate, distortion);
Jingning Han's avatar
Jingning Han committed
932

933
934
935
936
937
938
939
940
941
    if (this_rd < best_rd) {
      *bestrate = rate;
      *bestratey = ratey;
      *bestdistortion = distortion;
      best_rd = this_rd;
      *best_mode = mode;
      best_tx_type = tx_type;
      *a = tempa;
      *l = templ;
942
      vpx_memcpy(best_dqcoeff, BLOCK_OFFSET(xd->plane[0].dqcoeff, ib, 16), 32);
John Koleszar's avatar
John Koleszar committed
943
    }
John Koleszar's avatar
John Koleszar committed
944
  }
Scott LaVarnway's avatar
Scott LaVarnway committed
945
946
  xd->mode_info_context->bmi[ib].as_mode.first =
    (B_PREDICTION_MODE)(*best_mode);
947

Jingning Han's avatar
Jingning Han committed
948
  // inverse transform