summaryrefslogtreecommitdiffstats
path: root/third_party/aom/av1/encoder/rd.c
blob: 94c3bb96def2cd216d96da852e32c0b350cfb724 (plain)
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
26
27
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
107
108
109
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
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
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
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
706
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
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
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
950
951
952
953
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
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
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
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
1135
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
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
/*
 * Copyright (c) 2016, Alliance for Open Media. All rights reserved
 *
 * This source code is subject to the terms of the BSD 2 Clause License and
 * the Alliance for Open Media Patent License 1.0. If the BSD 2 Clause License
 * was not distributed with this source code in the LICENSE file, you can
 * obtain it at www.aomedia.org/license/software. If the Alliance for Open
 * Media Patent License 1.0 was not distributed with this source code in the
 * PATENTS file, you can obtain it at www.aomedia.org/license/patent.
 */

#include <assert.h>
#include <math.h>
#include <stdio.h>

#include "./av1_rtcd.h"

#include "aom_dsp/aom_dsp_common.h"
#include "aom_mem/aom_mem.h"
#include "aom_ports/bitops.h"
#include "aom_ports/mem.h"
#include "aom_ports/system_state.h"

#include "av1/common/common.h"
#include "av1/common/entropy.h"
#include "av1/common/entropymode.h"
#include "av1/common/mvref_common.h"
#include "av1/common/pred_common.h"
#include "av1/common/quant_common.h"
#include "av1/common/reconinter.h"
#include "av1/common/reconintra.h"
#include "av1/common/seg_common.h"

#include "av1/encoder/av1_quantize.h"
#include "av1/encoder/cost.h"
#include "av1/encoder/encodemb.h"
#include "av1/encoder/encodemv.h"
#include "av1/encoder/encoder.h"
#include "av1/encoder/mcomp.h"
#include "av1/encoder/ratectrl.h"
#include "av1/encoder/rd.h"
#include "av1/encoder/tokenize.h"

#define RD_THRESH_POW 1.25

// Factor to weigh the rate for switchable interp filters.
#define SWITCHABLE_INTERP_RATE_FACTOR 1

// The baseline rd thresholds for breaking out of the rd loop for
// certain modes are assumed to be based on 8x8 blocks.
// This table is used to correct for block size.
// The factors here are << 2 (2 = x0.5, 32 = x8 etc).
static const uint8_t rd_thresh_block_size_factor[BLOCK_SIZES] = {
#if CONFIG_CB4X4
  2,  2,  2,
#endif
  2,  3,  3, 4, 6, 6, 8, 12, 12, 16, 24, 24, 32,
#if CONFIG_EXT_PARTITION
  48, 48, 64
#endif  // CONFIG_EXT_PARTITION
};

static void fill_mode_costs(AV1_COMP *cpi) {
  const FRAME_CONTEXT *const fc = cpi->common.fc;
  int i, j;

  for (i = 0; i < INTRA_MODES; ++i)
    for (j = 0; j < INTRA_MODES; ++j)
      av1_cost_tokens(cpi->y_mode_costs[i][j], av1_kf_y_mode_prob[i][j],
                      av1_intra_mode_tree);

  for (i = 0; i < BLOCK_SIZE_GROUPS; ++i)
    av1_cost_tokens(cpi->mbmode_cost[i], fc->y_mode_prob[i],
                    av1_intra_mode_tree);

  for (i = 0; i < INTRA_MODES; ++i)
    av1_cost_tokens(cpi->intra_uv_mode_cost[i], fc->uv_mode_prob[i],
                    av1_intra_mode_tree);

  for (i = 0; i < SWITCHABLE_FILTER_CONTEXTS; ++i)
    av1_cost_tokens(cpi->switchable_interp_costs[i],
                    fc->switchable_interp_prob[i], av1_switchable_interp_tree);

#if CONFIG_PALETTE
  for (i = 0; i < PALETTE_BLOCK_SIZES; ++i) {
    av1_cost_tokens(cpi->palette_y_size_cost[i],
                    av1_default_palette_y_size_prob[i], av1_palette_size_tree);
    av1_cost_tokens(cpi->palette_uv_size_cost[i],
                    av1_default_palette_uv_size_prob[i], av1_palette_size_tree);
  }

  for (i = 0; i < PALETTE_SIZES; ++i) {
    for (j = 0; j < PALETTE_COLOR_INDEX_CONTEXTS; ++j) {
      av1_cost_tokens(cpi->palette_y_color_cost[i][j],
                      av1_default_palette_y_color_index_prob[i][j],
                      av1_palette_color_index_tree[i]);
      av1_cost_tokens(cpi->palette_uv_color_cost[i][j],
                      av1_default_palette_uv_color_index_prob[i][j],
                      av1_palette_color_index_tree[i]);
    }
  }
#endif  // CONFIG_PALETTE

  for (i = 0; i < MAX_TX_DEPTH; ++i)
    for (j = 0; j < TX_SIZE_CONTEXTS; ++j)
      av1_cost_tokens(cpi->tx_size_cost[i][j], fc->tx_size_probs[i][j],
                      av1_tx_size_tree[i]);

#if CONFIG_EXT_TX
  for (i = TX_4X4; i < EXT_TX_SIZES; ++i) {
    int s;
    for (s = 1; s < EXT_TX_SETS_INTER; ++s) {
      if (use_inter_ext_tx_for_txsize[s][i]) {
        av1_cost_tokens(cpi->inter_tx_type_costs[s][i],
                        fc->inter_ext_tx_prob[s][i], av1_ext_tx_inter_tree[s]);
      }
    }
    for (s = 1; s < EXT_TX_SETS_INTRA; ++s) {
      if (use_intra_ext_tx_for_txsize[s][i]) {
        for (j = 0; j < INTRA_MODES; ++j)
          av1_cost_tokens(cpi->intra_tx_type_costs[s][i][j],
                          fc->intra_ext_tx_prob[s][i][j],
                          av1_ext_tx_intra_tree[s]);
      }
    }
  }
#else
  for (i = TX_4X4; i < EXT_TX_SIZES; ++i) {
    for (j = 0; j < TX_TYPES; ++j)
      av1_cost_tokens(cpi->intra_tx_type_costs[i][j],
                      fc->intra_ext_tx_prob[i][j], av1_ext_tx_tree);
  }
  for (i = TX_4X4; i < EXT_TX_SIZES; ++i) {
    av1_cost_tokens(cpi->inter_tx_type_costs[i], fc->inter_ext_tx_prob[i],
                    av1_ext_tx_tree);
  }
#endif  // CONFIG_EXT_TX
#if CONFIG_EXT_INTRA
#if CONFIG_INTRA_INTERP
  for (i = 0; i < INTRA_FILTERS + 1; ++i)
    av1_cost_tokens(cpi->intra_filter_cost[i], fc->intra_filter_probs[i],
                    av1_intra_filter_tree);
#endif  // CONFIG_INTRA_INTERP
#endif  // CONFIG_EXT_INTRA
#if CONFIG_LOOP_RESTORATION
  av1_cost_tokens(cpi->switchable_restore_cost, fc->switchable_restore_prob,
                  av1_switchable_restore_tree);
#endif  // CONFIG_LOOP_RESTORATION
#if CONFIG_GLOBAL_MOTION
  av1_cost_tokens(cpi->gmtype_cost, fc->global_motion_types_prob,
                  av1_global_motion_types_tree);
#endif  // CONFIG_GLOBAL_MOTION
}

void av1_fill_token_costs(av1_coeff_cost *c,
                          av1_coeff_probs_model (*p)[PLANE_TYPES]) {
  int i, j, k, l;
  TX_SIZE t;
  for (t = 0; t < TX_SIZES; ++t)
    for (i = 0; i < PLANE_TYPES; ++i)
      for (j = 0; j < REF_TYPES; ++j)
        for (k = 0; k < COEF_BANDS; ++k)
          for (l = 0; l < BAND_COEFF_CONTEXTS(k); ++l) {
            aom_prob probs[ENTROPY_NODES];
            av1_model_to_full_probs(p[t][i][j][k][l], probs);
            av1_cost_tokens((int *)c[t][i][j][k][0][l], probs, av1_coef_tree);
            av1_cost_tokens_skip((int *)c[t][i][j][k][1][l], probs,
                                 av1_coef_tree);
            assert(c[t][i][j][k][0][l][EOB_TOKEN] ==
                   c[t][i][j][k][1][l][EOB_TOKEN]);
          }
}

// Values are now correlated to quantizer.
static int sad_per_bit16lut_8[QINDEX_RANGE];
static int sad_per_bit4lut_8[QINDEX_RANGE];

#if CONFIG_HIGHBITDEPTH
static int sad_per_bit16lut_10[QINDEX_RANGE];
static int sad_per_bit4lut_10[QINDEX_RANGE];
static int sad_per_bit16lut_12[QINDEX_RANGE];
static int sad_per_bit4lut_12[QINDEX_RANGE];
#endif

static void init_me_luts_bd(int *bit16lut, int *bit4lut, int range,
                            aom_bit_depth_t bit_depth) {
  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 < range; i++) {
    const double q = av1_convert_qindex_to_q(i, bit_depth);
    bit16lut[i] = (int)(0.0418 * q + 2.4107);
    bit4lut[i] = (int)(0.063 * q + 2.742);
  }
}

void av1_init_me_luts(void) {
  init_me_luts_bd(sad_per_bit16lut_8, sad_per_bit4lut_8, QINDEX_RANGE,
                  AOM_BITS_8);
#if CONFIG_HIGHBITDEPTH
  init_me_luts_bd(sad_per_bit16lut_10, sad_per_bit4lut_10, QINDEX_RANGE,
                  AOM_BITS_10);
  init_me_luts_bd(sad_per_bit16lut_12, sad_per_bit4lut_12, QINDEX_RANGE,
                  AOM_BITS_12);
#endif
}

static const int rd_boost_factor[16] = { 64, 32, 32, 32, 24, 16, 12, 12,
                                         8,  8,  4,  4,  2,  2,  1,  0 };
static const int rd_frame_type_factor[FRAME_UPDATE_TYPES] = {
  128, 144, 128, 128, 144,
#if CONFIG_EXT_REFS
  // TODO(zoeliu): To adjust further following factor values.
  128, 128, 128
  // TODO(weitinglin): We should investigate if the values should be the same
  //                   as the value used by OVERLAY frame
  ,
  144
#endif  // CONFIG_EXT_REFS
};

int av1_compute_rd_mult(const AV1_COMP *cpi, int qindex) {
  const int64_t q = av1_dc_quant(qindex, 0, cpi->common.bit_depth);
#if CONFIG_HIGHBITDEPTH
  int64_t rdmult = 0;
  switch (cpi->common.bit_depth) {
    case AOM_BITS_8: rdmult = 88 * q * q / 24; break;
    case AOM_BITS_10: rdmult = ROUND_POWER_OF_TWO(88 * q * q / 24, 4); break;
    case AOM_BITS_12: rdmult = ROUND_POWER_OF_TWO(88 * q * q / 24, 8); break;
    default:
      assert(0 && "bit_depth should be AOM_BITS_8, AOM_BITS_10 or AOM_BITS_12");
      return -1;
  }
#else
  int64_t rdmult = 88 * q * q / 24;
#endif  // CONFIG_HIGHBITDEPTH
  if (cpi->oxcf.pass == 2 && (cpi->common.frame_type != KEY_FRAME)) {
    const GF_GROUP *const gf_group = &cpi->twopass.gf_group;
    const FRAME_UPDATE_TYPE frame_type = gf_group->update_type[gf_group->index];
    const int boost_index = AOMMIN(15, (cpi->rc.gfu_boost / 100));

    rdmult = (rdmult * rd_frame_type_factor[frame_type]) >> 7;
    rdmult += ((rdmult * rd_boost_factor[boost_index]) >> 7);
  }
  if (rdmult < 1) rdmult = 1;
  return (int)rdmult;
}

static int compute_rd_thresh_factor(int qindex, aom_bit_depth_t bit_depth) {
  double q;
#if CONFIG_HIGHBITDEPTH
  switch (bit_depth) {
    case AOM_BITS_8: q = av1_dc_quant(qindex, 0, AOM_BITS_8) / 4.0; break;
    case AOM_BITS_10: q = av1_dc_quant(qindex, 0, AOM_BITS_10) / 16.0; break;
    case AOM_BITS_12: q = av1_dc_quant(qindex, 0, AOM_BITS_12) / 64.0; break;
    default:
      assert(0 && "bit_depth should be AOM_BITS_8, AOM_BITS_10 or AOM_BITS_12");
      return -1;
  }
#else
  (void)bit_depth;
  q = av1_dc_quant(qindex, 0, AOM_BITS_8) / 4.0;
#endif  // CONFIG_HIGHBITDEPTH
  // TODO(debargha): Adjust the function below.
  return AOMMAX((int)(pow(q, RD_THRESH_POW) * 5.12), 8);
}

void av1_initialize_me_consts(const AV1_COMP *cpi, MACROBLOCK *x, int qindex) {
#if CONFIG_HIGHBITDEPTH
  switch (cpi->common.bit_depth) {
    case AOM_BITS_8:
      x->sadperbit16 = sad_per_bit16lut_8[qindex];
      x->sadperbit4 = sad_per_bit4lut_8[qindex];
      break;
    case AOM_BITS_10:
      x->sadperbit16 = sad_per_bit16lut_10[qindex];
      x->sadperbit4 = sad_per_bit4lut_10[qindex];
      break;
    case AOM_BITS_12:
      x->sadperbit16 = sad_per_bit16lut_12[qindex];
      x->sadperbit4 = sad_per_bit4lut_12[qindex];
      break;
    default:
      assert(0 && "bit_depth should be AOM_BITS_8, AOM_BITS_10 or AOM_BITS_12");
  }
#else
  (void)cpi;
  x->sadperbit16 = sad_per_bit16lut_8[qindex];
  x->sadperbit4 = sad_per_bit4lut_8[qindex];
#endif  // CONFIG_HIGHBITDEPTH
}

static void set_block_thresholds(const AV1_COMMON *cm, RD_OPT *rd) {
  int i, bsize, segment_id;

  for (segment_id = 0; segment_id < MAX_SEGMENTS; ++segment_id) {
    const int qindex =
        clamp(av1_get_qindex(&cm->seg, segment_id, cm->base_qindex) +
                  cm->y_dc_delta_q,
              0, MAXQ);
    const int q = compute_rd_thresh_factor(qindex, cm->bit_depth);

    for (bsize = 0; bsize < BLOCK_SIZES; ++bsize) {
      // Threshold here seems unnecessarily harsh but fine given actual
      // range of values used for cpi->sf.thresh_mult[].
      const int t = q * rd_thresh_block_size_factor[bsize];
      const int thresh_max = INT_MAX / t;

#if CONFIG_CB4X4
      for (i = 0; i < MAX_MODES; ++i)
        rd->threshes[segment_id][bsize][i] = rd->thresh_mult[i] < thresh_max
                                                 ? rd->thresh_mult[i] * t / 4
                                                 : INT_MAX;
#else
      if (bsize >= BLOCK_8X8) {
        for (i = 0; i < MAX_MODES; ++i)
          rd->threshes[segment_id][bsize][i] = rd->thresh_mult[i] < thresh_max
                                                   ? rd->thresh_mult[i] * t / 4
                                                   : INT_MAX;
      } else {
        for (i = 0; i < MAX_REFS; ++i)
          rd->threshes[segment_id][bsize][i] =
              rd->thresh_mult_sub8x8[i] < thresh_max
                  ? rd->thresh_mult_sub8x8[i] * t / 4
                  : INT_MAX;
      }
#endif
    }
  }
}

void av1_set_mvcost(MACROBLOCK *x, MV_REFERENCE_FRAME ref_frame, int ref,
                    int ref_mv_idx) {
  MB_MODE_INFO_EXT *mbmi_ext = x->mbmi_ext;
  int8_t rf_type = av1_ref_frame_type(x->e_mbd.mi[0]->mbmi.ref_frame);
  int nmv_ctx = av1_nmv_ctx(mbmi_ext->ref_mv_count[rf_type],
                            mbmi_ext->ref_mv_stack[rf_type], ref, ref_mv_idx);
  (void)ref_frame;
  x->mvcost = x->mv_cost_stack[nmv_ctx];
  x->nmvjointcost = x->nmv_vec_cost[nmv_ctx];
}

void av1_initialize_rd_consts(AV1_COMP *cpi) {
  AV1_COMMON *const cm = &cpi->common;
  MACROBLOCK *const x = &cpi->td.mb;
  RD_OPT *const rd = &cpi->rd;
  int i;
  int nmv_ctx;

  aom_clear_system_state();

  rd->RDDIV = RDDIV_BITS;  // In bits (to multiply D by 128).
  rd->RDMULT = av1_compute_rd_mult(cpi, cm->base_qindex + cm->y_dc_delta_q);

  set_error_per_bit(x, rd->RDMULT);

  set_block_thresholds(cm, rd);

  for (nmv_ctx = 0; nmv_ctx < NMV_CONTEXTS; ++nmv_ctx) {
    av1_build_nmv_cost_table(
        x->nmv_vec_cost[nmv_ctx],
        cm->allow_high_precision_mv ? x->nmvcost_hp[nmv_ctx]
                                    : x->nmvcost[nmv_ctx],
        &cm->fc->nmvc[nmv_ctx], cm->allow_high_precision_mv);
  }
  x->mvcost = x->mv_cost_stack[0];
  x->nmvjointcost = x->nmv_vec_cost[0];

  if (cpi->oxcf.pass != 1) {
    av1_fill_token_costs(x->token_costs, cm->fc->coef_probs);

    if (cm->frame_type == KEY_FRAME) {
#if CONFIG_EXT_PARTITION_TYPES
      for (i = 0; i < PARTITION_PLOFFSET; ++i)
        av1_cost_tokens(cpi->partition_cost[i], cm->fc->partition_prob[i],
                        av1_partition_tree);
      for (; i < PARTITION_CONTEXTS_PRIMARY; ++i)
        av1_cost_tokens(cpi->partition_cost[i], cm->fc->partition_prob[i],
                        av1_ext_partition_tree);
#else
      for (i = 0; i < PARTITION_CONTEXTS_PRIMARY; ++i)
        av1_cost_tokens(cpi->partition_cost[i], cm->fc->partition_prob[i],
                        av1_partition_tree);
#endif  // CONFIG_EXT_PARTITION_TYPES
#if CONFIG_UNPOISON_PARTITION_CTX
      for (; i < PARTITION_CONTEXTS_PRIMARY + PARTITION_BLOCK_SIZES; ++i) {
        aom_prob p = cm->fc->partition_prob[i][PARTITION_VERT];
        assert(p > 0);
        cpi->partition_cost[i][PARTITION_NONE] = INT_MAX;
        cpi->partition_cost[i][PARTITION_HORZ] = INT_MAX;
        cpi->partition_cost[i][PARTITION_VERT] = av1_cost_bit(p, 0);
        cpi->partition_cost[i][PARTITION_SPLIT] = av1_cost_bit(p, 1);
      }
      for (; i < PARTITION_CONTEXTS_PRIMARY + 2 * PARTITION_BLOCK_SIZES; ++i) {
        aom_prob p = cm->fc->partition_prob[i][PARTITION_HORZ];
        assert(p > 0);
        cpi->partition_cost[i][PARTITION_NONE] = INT_MAX;
        cpi->partition_cost[i][PARTITION_HORZ] = av1_cost_bit(p, 0);
        cpi->partition_cost[i][PARTITION_VERT] = INT_MAX;
        cpi->partition_cost[i][PARTITION_SPLIT] = av1_cost_bit(p, 1);
      }
      cpi->partition_cost[PARTITION_CONTEXTS][PARTITION_NONE] = INT_MAX;
      cpi->partition_cost[PARTITION_CONTEXTS][PARTITION_HORZ] = INT_MAX;
      cpi->partition_cost[PARTITION_CONTEXTS][PARTITION_VERT] = INT_MAX;
      cpi->partition_cost[PARTITION_CONTEXTS][PARTITION_SPLIT] = 0;
#endif  // CONFIG_UNPOISON_PARTITION_CTX
    }

    fill_mode_costs(cpi);

    if (!frame_is_intra_only(cm)) {
      for (i = 0; i < NEWMV_MODE_CONTEXTS; ++i) {
        cpi->newmv_mode_cost[i][0] = av1_cost_bit(cm->fc->newmv_prob[i], 0);
        cpi->newmv_mode_cost[i][1] = av1_cost_bit(cm->fc->newmv_prob[i], 1);
      }

      for (i = 0; i < ZEROMV_MODE_CONTEXTS; ++i) {
        cpi->zeromv_mode_cost[i][0] = av1_cost_bit(cm->fc->zeromv_prob[i], 0);
        cpi->zeromv_mode_cost[i][1] = av1_cost_bit(cm->fc->zeromv_prob[i], 1);
      }

      for (i = 0; i < REFMV_MODE_CONTEXTS; ++i) {
        cpi->refmv_mode_cost[i][0] = av1_cost_bit(cm->fc->refmv_prob[i], 0);
        cpi->refmv_mode_cost[i][1] = av1_cost_bit(cm->fc->refmv_prob[i], 1);
      }

      for (i = 0; i < DRL_MODE_CONTEXTS; ++i) {
        cpi->drl_mode_cost0[i][0] = av1_cost_bit(cm->fc->drl_prob[i], 0);
        cpi->drl_mode_cost0[i][1] = av1_cost_bit(cm->fc->drl_prob[i], 1);
      }
#if CONFIG_EXT_INTER
      for (i = 0; i < INTER_MODE_CONTEXTS; ++i)
        av1_cost_tokens((int *)cpi->inter_compound_mode_cost[i],
                        cm->fc->inter_compound_mode_probs[i],
                        av1_inter_compound_mode_tree);
#if CONFIG_INTERINTRA
      for (i = 0; i < BLOCK_SIZE_GROUPS; ++i)
        av1_cost_tokens((int *)cpi->interintra_mode_cost[i],
                        cm->fc->interintra_mode_prob[i],
                        av1_interintra_mode_tree);
#endif  // CONFIG_INTERINTRA
#endif  // CONFIG_EXT_INTER
#if CONFIG_MOTION_VAR || CONFIG_WARPED_MOTION
      for (i = BLOCK_8X8; i < BLOCK_SIZES; i++) {
        av1_cost_tokens((int *)cpi->motion_mode_cost[i],
                        cm->fc->motion_mode_prob[i], av1_motion_mode_tree);
      }
#if CONFIG_MOTION_VAR && CONFIG_WARPED_MOTION
      for (i = BLOCK_8X8; i < BLOCK_SIZES; i++) {
        cpi->motion_mode_cost1[i][0] = av1_cost_bit(cm->fc->obmc_prob[i], 0);
        cpi->motion_mode_cost1[i][1] = av1_cost_bit(cm->fc->obmc_prob[i], 1);
      }
#endif  // CONFIG_MOTION_VAR && CONFIG_WARPED_MOTION
#endif  // CONFIG_MOTION_VAR || CONFIG_WARPED_MOTION
    }
  }
}

static void model_rd_norm(int xsq_q10, int *r_q10, int *d_q10) {
  // NOTE: The tables below must be of the same size.

  // The functions described below are sampled at the four most significant
  // bits of x^2 + 8 / 256.

  // Normalized rate:
  // This table models the rate for a Laplacian source with given variance
  // when quantized with a uniform quantizer with given stepsize. The
  // closed form expression is:
  // Rn(x) = H(sqrt(r)) + sqrt(r)*[1 + H(r)/(1 - r)],
  // where r = exp(-sqrt(2) * x) and x = qpstep / sqrt(variance),
  // and H(x) is the binary entropy function.
  static const int rate_tab_q10[] = {
    65536, 6086, 5574, 5275, 5063, 4899, 4764, 4651, 4553, 4389, 4255, 4142,
    4044,  3958, 3881, 3811, 3748, 3635, 3538, 3453, 3376, 3307, 3244, 3186,
    3133,  3037, 2952, 2877, 2809, 2747, 2690, 2638, 2589, 2501, 2423, 2353,
    2290,  2232, 2179, 2130, 2084, 2001, 1928, 1862, 1802, 1748, 1698, 1651,
    1608,  1530, 1460, 1398, 1342, 1290, 1243, 1199, 1159, 1086, 1021, 963,
    911,   864,  821,  781,  745,  680,  623,  574,  530,  490,  455,  424,
    395,   345,  304,  269,  239,  213,  190,  171,  154,  126,  104,  87,
    73,    61,   52,   44,   38,   28,   21,   16,   12,   10,   8,    6,
    5,     3,    2,    1,    1,    1,    0,    0,
  };
  // Normalized distortion:
  // This table models the normalized distortion for a Laplacian source
  // with given variance when quantized with a uniform quantizer
  // with given stepsize. The closed form expression is:
  // Dn(x) = 1 - 1/sqrt(2) * x / sinh(x/sqrt(2))
  // where x = qpstep / sqrt(variance).
  // Note the actual distortion is Dn * variance.
  static const int dist_tab_q10[] = {
    0,    0,    1,    1,    1,    2,    2,    2,    3,    3,    4,    5,
    5,    6,    7,    7,    8,    9,    11,   12,   13,   15,   16,   17,
    18,   21,   24,   26,   29,   31,   34,   36,   39,   44,   49,   54,
    59,   64,   69,   73,   78,   88,   97,   106,  115,  124,  133,  142,
    151,  167,  184,  200,  215,  231,  245,  260,  274,  301,  327,  351,
    375,  397,  418,  439,  458,  495,  528,  559,  587,  613,  637,  659,
    680,  717,  749,  777,  801,  823,  842,  859,  874,  899,  919,  936,
    949,  960,  969,  977,  983,  994,  1001, 1006, 1010, 1013, 1015, 1017,
    1018, 1020, 1022, 1022, 1023, 1023, 1023, 1024,
  };
  static const int xsq_iq_q10[] = {
    0,      4,      8,      12,     16,     20,     24,     28,     32,
    40,     48,     56,     64,     72,     80,     88,     96,     112,
    128,    144,    160,    176,    192,    208,    224,    256,    288,
    320,    352,    384,    416,    448,    480,    544,    608,    672,
    736,    800,    864,    928,    992,    1120,   1248,   1376,   1504,
    1632,   1760,   1888,   2016,   2272,   2528,   2784,   3040,   3296,
    3552,   3808,   4064,   4576,   5088,   5600,   6112,   6624,   7136,
    7648,   8160,   9184,   10208,  11232,  12256,  13280,  14304,  15328,
    16352,  18400,  20448,  22496,  24544,  26592,  28640,  30688,  32736,
    36832,  40928,  45024,  49120,  53216,  57312,  61408,  65504,  73696,
    81888,  90080,  98272,  106464, 114656, 122848, 131040, 147424, 163808,
    180192, 196576, 212960, 229344, 245728,
  };
  const int tmp = (xsq_q10 >> 2) + 8;
  const int k = get_msb(tmp) - 3;
  const int xq = (k << 3) + ((tmp >> k) & 0x7);
  const int one_q10 = 1 << 10;
  const int a_q10 = ((xsq_q10 - xsq_iq_q10[xq]) << 10) >> (2 + k);
  const int b_q10 = one_q10 - a_q10;
  *r_q10 = (rate_tab_q10[xq] * b_q10 + rate_tab_q10[xq + 1] * a_q10) >> 10;
  *d_q10 = (dist_tab_q10[xq] * b_q10 + dist_tab_q10[xq + 1] * a_q10) >> 10;
}

void av1_model_rd_from_var_lapndz(int64_t var, unsigned int n_log2,
                                  unsigned int qstep, int *rate,
                                  int64_t *dist) {
  // This function models the rate and distortion for a Laplacian
  // source with given variance when quantized with a uniform quantizer
  // with given stepsize. The closed form expressions are in:
  // Hang and Chen, "Source Model for transform video coder and its
  // application - Part I: Fundamental Theory", IEEE Trans. Circ.
  // Sys. for Video Tech., April 1997.
  if (var == 0) {
    *rate = 0;
    *dist = 0;
  } else {
    int d_q10, r_q10;
    static const uint32_t MAX_XSQ_Q10 = 245727;
    const uint64_t xsq_q10_64 =
        (((uint64_t)qstep * qstep << (n_log2 + 10)) + (var >> 1)) / var;
    const int xsq_q10 = (int)AOMMIN(xsq_q10_64, MAX_XSQ_Q10);
    model_rd_norm(xsq_q10, &r_q10, &d_q10);
    *rate = ROUND_POWER_OF_TWO(r_q10 << n_log2, 10 - AV1_PROB_COST_SHIFT);
    *dist = (var * (int64_t)d_q10 + 512) >> 10;
  }
}

static void get_entropy_contexts_plane(
    BLOCK_SIZE plane_bsize, TX_SIZE tx_size, const struct macroblockd_plane *pd,
    ENTROPY_CONTEXT t_above[2 * MAX_MIB_SIZE],
    ENTROPY_CONTEXT t_left[2 * MAX_MIB_SIZE]) {
  const int num_4x4_w = block_size_wide[plane_bsize] >> tx_size_wide_log2[0];
  const int num_4x4_h = block_size_high[plane_bsize] >> tx_size_high_log2[0];
  const ENTROPY_CONTEXT *const above = pd->above_context;
  const ENTROPY_CONTEXT *const left = pd->left_context;

#if CONFIG_LV_MAP
  memcpy(t_above, above, sizeof(ENTROPY_CONTEXT) * num_4x4_w);
  memcpy(t_left, left, sizeof(ENTROPY_CONTEXT) * num_4x4_h);
  return;
#endif  // CONFIG_LV_MAP

  int i;

#if CONFIG_CHROMA_2X2
  switch (tx_size) {
    case TX_2X2:
      memcpy(t_above, above, sizeof(ENTROPY_CONTEXT) * num_4x4_w);
      memcpy(t_left, left, sizeof(ENTROPY_CONTEXT) * num_4x4_h);
      break;
    case TX_4X4:
      for (i = 0; i < num_4x4_w; i += 2)
        t_above[i] = !!*(const uint16_t *)&above[i];
      for (i = 0; i < num_4x4_h; i += 2)
        t_left[i] = !!*(const uint16_t *)&left[i];
      break;
    case TX_8X8:
      for (i = 0; i < num_4x4_w; i += 4)
        t_above[i] = !!*(const uint32_t *)&above[i];
      for (i = 0; i < num_4x4_h; i += 4)
        t_left[i] = !!*(const uint32_t *)&left[i];
      break;
    case TX_16X16:
      for (i = 0; i < num_4x4_w; i += 8)
        t_above[i] = !!*(const uint64_t *)&above[i];
      for (i = 0; i < num_4x4_h; i += 8)
        t_left[i] = !!*(const uint64_t *)&left[i];
      break;
    case TX_32X32:
      for (i = 0; i < num_4x4_w; i += 16)
        t_above[i] =
            !!(*(const uint64_t *)&above[i] | *(const uint64_t *)&above[i + 8]);
      for (i = 0; i < num_4x4_h; i += 16)
        t_left[i] =
            !!(*(const uint64_t *)&left[i] | *(const uint64_t *)&left[i + 8]);
      break;
#if CONFIG_TX64X64
    case TX_64X64:
      for (i = 0; i < num_4x4_w; i += 32)
        t_above[i] =
            !!(*(const uint64_t *)&above[i] | *(const uint64_t *)&above[i + 8] |
               *(const uint64_t *)&above[i + 16] |
               *(const uint64_t *)&above[i + 24]);
      for (i = 0; i < num_4x4_h; i += 32)
        t_left[i] =
            !!(*(const uint64_t *)&left[i] | *(const uint64_t *)&left[i + 8] |
               *(const uint64_t *)&left[i + 16] |
               *(const uint64_t *)&left[i + 24]);
      break;
#endif  // CONFIG_TX64X64
    case TX_4X8:
      for (i = 0; i < num_4x4_w; i += 2)
        t_above[i] = !!*(const uint16_t *)&above[i];
      for (i = 0; i < num_4x4_h; i += 4)
        t_left[i] = !!*(const uint32_t *)&left[i];
      break;
    case TX_8X4:
      for (i = 0; i < num_4x4_w; i += 4)
        t_above[i] = !!*(const uint32_t *)&above[i];
      for (i = 0; i < num_4x4_h; i += 2)
        t_left[i] = !!*(const uint16_t *)&left[i];
      break;
    case TX_8X16:
      for (i = 0; i < num_4x4_w; i += 4)
        t_above[i] = !!*(const uint32_t *)&above[i];
      for (i = 0; i < num_4x4_h; i += 8)
        t_left[i] = !!*(const uint64_t *)&left[i];
      break;
    case TX_16X8:
      for (i = 0; i < num_4x4_w; i += 8)
        t_above[i] = !!*(const uint64_t *)&above[i];
      for (i = 0; i < num_4x4_h; i += 4)
        t_left[i] = !!*(const uint32_t *)&left[i];
      break;
    case TX_16X32:
      for (i = 0; i < num_4x4_w; i += 8)
        t_above[i] = !!*(const uint64_t *)&above[i];
      for (i = 0; i < num_4x4_h; i += 16)
        t_left[i] =
            !!(*(const uint64_t *)&left[i] | *(const uint64_t *)&left[i + 8]);
      break;
    case TX_32X16:
      for (i = 0; i < num_4x4_w; i += 16)
        t_above[i] =
            !!(*(const uint64_t *)&above[i] | *(const uint64_t *)&above[i + 8]);
      for (i = 0; i < num_4x4_h; i += 8)
        t_left[i] = !!*(const uint64_t *)&left[i];
      break;
#if CONFIG_EXT_TX && CONFIG_RECT_TX && CONFIG_RECT_TX_EXT
    case TX_4X16:
      for (i = 0; i < num_4x4_w; i += 2)
        t_above[i] = !!*(const uint16_t *)&above[i];
      for (i = 0; i < num_4x4_h; i += 8)
        t_left[i] = !!*(const uint64_t *)&left[i];
      break;
    case TX_16X4:
      for (i = 0; i < num_4x4_w; i += 8)
        t_above[i] = !!*(const uint64_t *)&above[i];
      for (i = 0; i < num_4x4_h; i += 2)
        t_left[i] = !!*(const uint16_t *)&left[i];
      break;
    case TX_8X32:
      for (i = 0; i < num_4x4_w; i += 4)
        t_above[i] = !!*(const uint32_t *)&above[i];
      for (i = 0; i < num_4x4_h; i += 16)
        t_left[i] =
            !!(*(const uint64_t *)&left[i] | *(const uint64_t *)&left[i + 8]);
      break;
    case TX_32X8:
      for (i = 0; i < num_4x4_w; i += 16)
        t_above[i] =
            !!(*(const uint64_t *)&above[i] | *(const uint64_t *)&above[i + 8]);
      for (i = 0; i < num_4x4_h; i += 4)
        t_left[i] = !!*(const uint32_t *)&left[i];
      break;
#endif  // CONFIG_EXT_TX && CONFIG_RECT_TX && CONFIG_RECT_TX_EXT

    default: assert(0 && "Invalid transform size."); break;
  }
  return;
#endif  // CONFIG_CHROMA_2X2

  switch (tx_size) {
    case TX_4X4:
      memcpy(t_above, above, sizeof(ENTROPY_CONTEXT) * num_4x4_w);
      memcpy(t_left, left, sizeof(ENTROPY_CONTEXT) * num_4x4_h);
      break;
    case TX_8X8:
      for (i = 0; i < num_4x4_w; i += 2)
        t_above[i] = !!*(const uint16_t *)&above[i];
      for (i = 0; i < num_4x4_h; i += 2)
        t_left[i] = !!*(const uint16_t *)&left[i];
      break;
    case TX_16X16:
      for (i = 0; i < num_4x4_w; i += 4)
        t_above[i] = !!*(const uint32_t *)&above[i];
      for (i = 0; i < num_4x4_h; i += 4)
        t_left[i] = !!*(const uint32_t *)&left[i];
      break;
    case TX_32X32:
      for (i = 0; i < num_4x4_w; i += 8)
        t_above[i] = !!*(const uint64_t *)&above[i];
      for (i = 0; i < num_4x4_h; i += 8)
        t_left[i] = !!*(const uint64_t *)&left[i];
      break;
#if CONFIG_TX64X64
    case TX_64X64:
      for (i = 0; i < num_4x4_w; i += 16)
        t_above[i] =
            !!(*(const uint64_t *)&above[i] | *(const uint64_t *)&above[i + 8]);
      for (i = 0; i < num_4x4_h; i += 16)
        t_left[i] =
            !!(*(const uint64_t *)&left[i] | *(const uint64_t *)&left[i + 8]);
      break;
#endif  // CONFIG_TX64X64
    case TX_4X8:
      memcpy(t_above, above, sizeof(ENTROPY_CONTEXT) * num_4x4_w);
      for (i = 0; i < num_4x4_h; i += 2)
        t_left[i] = !!*(const uint16_t *)&left[i];
      break;
    case TX_8X4:
      for (i = 0; i < num_4x4_w; i += 2)
        t_above[i] = !!*(const uint16_t *)&above[i];
      memcpy(t_left, left, sizeof(ENTROPY_CONTEXT) * num_4x4_h);
      break;
    case TX_8X16:
      for (i = 0; i < num_4x4_w; i += 2)
        t_above[i] = !!*(const uint16_t *)&above[i];
      for (i = 0; i < num_4x4_h; i += 4)
        t_left[i] = !!*(const uint32_t *)&left[i];
      break;
    case TX_16X8:
      for (i = 0; i < num_4x4_w; i += 4)
        t_above[i] = !!*(const uint32_t *)&above[i];
      for (i = 0; i < num_4x4_h; i += 2)
        t_left[i] = !!*(const uint16_t *)&left[i];
      break;
    case TX_16X32:
      for (i = 0; i < num_4x4_w; i += 4)
        t_above[i] = !!*(const uint32_t *)&above[i];
      for (i = 0; i < num_4x4_h; i += 8)
        t_left[i] = !!*(const uint64_t *)&left[i];
      break;
    case TX_32X16:
      for (i = 0; i < num_4x4_w; i += 8)
        t_above[i] = !!*(const uint64_t *)&above[i];
      for (i = 0; i < num_4x4_h; i += 4)
        t_left[i] = !!*(const uint32_t *)&left[i];
      break;
#if CONFIG_EXT_TX && CONFIG_RECT_TX && CONFIG_RECT_TX_EXT
    case TX_4X16:
      memcpy(t_above, above, sizeof(ENTROPY_CONTEXT) * num_4x4_w);
      for (i = 0; i < num_4x4_h; i += 4)
        t_left[i] = !!*(const uint32_t *)&left[i];
      break;
    case TX_16X4:
      for (i = 0; i < num_4x4_w; i += 4)
        t_above[i] = !!*(const uint32_t *)&above[i];
      memcpy(t_left, left, sizeof(ENTROPY_CONTEXT) * num_4x4_h);
      break;
    case TX_8X32:
      for (i = 0; i < num_4x4_w; i += 2)
        t_above[i] = !!*(const uint16_t *)&above[i];
      for (i = 0; i < num_4x4_h; i += 8)
        t_left[i] = !!*(const uint64_t *)&left[i];
      break;
    case TX_32X8:
      for (i = 0; i < num_4x4_w; i += 8)
        t_above[i] = !!*(const uint64_t *)&above[i];
      for (i = 0; i < num_4x4_h; i += 2)
        t_left[i] = !!*(const uint16_t *)&left[i];
      break;
#endif  // CONFIG_EXT_TX && CONFIG_RECT_TX && CONFIG_RECT_TX_EXT
    default: assert(0 && "Invalid transform size."); break;
  }
}

void av1_get_entropy_contexts(BLOCK_SIZE bsize, TX_SIZE tx_size,
                              const struct macroblockd_plane *pd,
                              ENTROPY_CONTEXT t_above[2 * MAX_MIB_SIZE],
                              ENTROPY_CONTEXT t_left[2 * MAX_MIB_SIZE]) {
#if CONFIG_CB4X4 && !CONFIG_CHROMA_2X2
  const BLOCK_SIZE plane_bsize =
      AOMMAX(BLOCK_4X4, get_plane_block_size(bsize, pd));
#else
  const BLOCK_SIZE plane_bsize = get_plane_block_size(bsize, pd);
#endif
  get_entropy_contexts_plane(plane_bsize, tx_size, pd, t_above, t_left);
}

void av1_mv_pred(const AV1_COMP *cpi, MACROBLOCK *x, uint8_t *ref_y_buffer,
                 int ref_y_stride, int ref_frame, BLOCK_SIZE block_size) {
  int i;
  int zero_seen = 0;
  int best_index = 0;
  int best_sad = INT_MAX;
  int this_sad = INT_MAX;
  int max_mv = 0;
  uint8_t *src_y_ptr = x->plane[0].src.buf;
  uint8_t *ref_y_ptr;
  MV pred_mv[MAX_MV_REF_CANDIDATES + 1];
  int num_mv_refs = 0;

  pred_mv[num_mv_refs++] = x->mbmi_ext->ref_mvs[ref_frame][0].as_mv;
  if (x->mbmi_ext->ref_mvs[ref_frame][0].as_int !=
      x->mbmi_ext->ref_mvs[ref_frame][1].as_int) {
    pred_mv[num_mv_refs++] = x->mbmi_ext->ref_mvs[ref_frame][1].as_mv;
  }
  if (cpi->sf.adaptive_motion_search && block_size < x->max_partition_size)
    pred_mv[num_mv_refs++] = x->pred_mv[ref_frame];

  assert(num_mv_refs <= (int)(sizeof(pred_mv) / sizeof(pred_mv[0])));

  // Get the sad for each candidate reference mv.
  for (i = 0; i < num_mv_refs; ++i) {
    const MV *this_mv = &pred_mv[i];
    int fp_row, fp_col;
    fp_row = (this_mv->row + 3 + (this_mv->row >= 0)) >> 3;
    fp_col = (this_mv->col + 3 + (this_mv->col >= 0)) >> 3;
    max_mv = AOMMAX(max_mv, AOMMAX(abs(this_mv->row), abs(this_mv->col)) >> 3);

    if (fp_row == 0 && fp_col == 0 && zero_seen) continue;
    zero_seen |= (fp_row == 0 && fp_col == 0);

    ref_y_ptr = &ref_y_buffer[ref_y_stride * fp_row + fp_col];
    // Find sad for current vector.
    this_sad = cpi->fn_ptr[block_size].sdf(src_y_ptr, x->plane[0].src.stride,
                                           ref_y_ptr, ref_y_stride);
    // Note if it is the best so far.
    if (this_sad < best_sad) {
      best_sad = this_sad;
      best_index = i;
    }
  }

  // Note the index of the mv that worked best in the reference list.
  x->mv_best_ref_index[ref_frame] = best_index;
  x->max_mv_context[ref_frame] = max_mv;
  x->pred_mv_sad[ref_frame] = best_sad;
}

void av1_setup_pred_block(const MACROBLOCKD *xd,
                          struct buf_2d dst[MAX_MB_PLANE],
                          const YV12_BUFFER_CONFIG *src, int mi_row, int mi_col,
                          const struct scale_factors *scale,
                          const struct scale_factors *scale_uv) {
  int i;

  dst[0].buf = src->y_buffer;
  dst[0].stride = src->y_stride;
  dst[1].buf = src->u_buffer;
  dst[2].buf = src->v_buffer;
  dst[1].stride = dst[2].stride = src->uv_stride;

  for (i = 0; i < MAX_MB_PLANE; ++i) {
    setup_pred_plane(dst + i, xd->mi[0]->mbmi.sb_type, dst[i].buf,
                     i ? src->uv_crop_width : src->y_crop_width,
                     i ? src->uv_crop_height : src->y_crop_height,
                     dst[i].stride, mi_row, mi_col, i ? scale_uv : scale,
                     xd->plane[i].subsampling_x, xd->plane[i].subsampling_y);
  }
}

int av1_raster_block_offset(BLOCK_SIZE plane_bsize, int raster_block,
                            int stride) {
  const int bw = b_width_log2_lookup[plane_bsize];
  const int y = 4 * (raster_block >> bw);
  const int x = 4 * (raster_block & ((1 << bw) - 1));
  return y * stride + x;
}

int16_t *av1_raster_block_offset_int16(BLOCK_SIZE plane_bsize, int raster_block,
                                       int16_t *base) {
  const int stride = block_size_wide[plane_bsize];
  return base + av1_raster_block_offset(plane_bsize, raster_block, stride);
}

YV12_BUFFER_CONFIG *av1_get_scaled_ref_frame(const AV1_COMP *cpi,
                                             int ref_frame) {
  const AV1_COMMON *const cm = &cpi->common;
  const int scaled_idx = cpi->scaled_ref_idx[ref_frame - 1];
  const int ref_idx = get_ref_frame_buf_idx(cpi, ref_frame);
  return (scaled_idx != ref_idx && scaled_idx != INVALID_IDX)
             ? &cm->buffer_pool->frame_bufs[scaled_idx].buf
             : NULL;
}

#if CONFIG_DUAL_FILTER
int av1_get_switchable_rate(const AV1_COMP *cpi, const MACROBLOCKD *xd) {
  const AV1_COMMON *const cm = &cpi->common;
  if (cm->interp_filter == SWITCHABLE) {
    const MB_MODE_INFO *const mbmi = &xd->mi[0]->mbmi;
    int inter_filter_cost = 0;
    int dir;

    for (dir = 0; dir < 2; ++dir) {
      if (has_subpel_mv_component(xd->mi[0], xd, dir) ||
          (mbmi->ref_frame[1] > INTRA_FRAME &&
           has_subpel_mv_component(xd->mi[0], xd, dir + 2))) {
        const int ctx = av1_get_pred_context_switchable_interp(xd, dir);
        inter_filter_cost +=
            cpi->switchable_interp_costs[ctx][mbmi->interp_filter[dir]];
      }
    }
    return SWITCHABLE_INTERP_RATE_FACTOR * inter_filter_cost;
  } else {
    return 0;
  }
}
#else
int av1_get_switchable_rate(const AV1_COMP *cpi, const MACROBLOCKD *xd) {
  const AV1_COMMON *const cm = &cpi->common;
  if (cm->interp_filter == SWITCHABLE) {
    const MB_MODE_INFO *const mbmi = &xd->mi[0]->mbmi;
    const int ctx = av1_get_pred_context_switchable_interp(xd);
    return SWITCHABLE_INTERP_RATE_FACTOR *
           cpi->switchable_interp_costs[ctx][mbmi->interp_filter];
  }
  return 0;
}
#endif

void av1_set_rd_speed_thresholds(AV1_COMP *cpi) {
  int i;
  RD_OPT *const rd = &cpi->rd;
  SPEED_FEATURES *const sf = &cpi->sf;

  // Set baseline threshold values.
  for (i = 0; i < MAX_MODES; ++i) rd->thresh_mult[i] = cpi->oxcf.mode == 0;

  if (sf->adaptive_rd_thresh) {
    rd->thresh_mult[THR_NEARESTMV] = 300;
#if CONFIG_EXT_REFS
    rd->thresh_mult[THR_NEARESTL2] = 300;
    rd->thresh_mult[THR_NEARESTL3] = 300;
    rd->thresh_mult[THR_NEARESTB] = 300;
#endif  // CONFIG_EXT_REFS
    rd->thresh_mult[THR_NEARESTA] = 300;
    rd->thresh_mult[THR_NEARESTG] = 300;
  } else {
    rd->thresh_mult[THR_NEARESTMV] = 0;
#if CONFIG_EXT_REFS
    rd->thresh_mult[THR_NEARESTL2] = 0;
    rd->thresh_mult[THR_NEARESTL3] = 0;
    rd->thresh_mult[THR_NEARESTB] = 0;
#endif  // CONFIG_EXT_REFS
    rd->thresh_mult[THR_NEARESTA] = 0;
    rd->thresh_mult[THR_NEARESTG] = 0;
  }

  rd->thresh_mult[THR_DC] += 1000;

  rd->thresh_mult[THR_NEWMV] += 1000;
#if CONFIG_EXT_REFS
  rd->thresh_mult[THR_NEWL2] += 1000;
  rd->thresh_mult[THR_NEWL3] += 1000;
  rd->thresh_mult[THR_NEWB] += 1000;
#endif  // CONFIG_EXT_REFS
  rd->thresh_mult[THR_NEWA] += 1000;
  rd->thresh_mult[THR_NEWG] += 1000;

  rd->thresh_mult[THR_NEARMV] += 1000;
#if CONFIG_EXT_REFS
  rd->thresh_mult[THR_NEARL2] += 1000;
  rd->thresh_mult[THR_NEARL3] += 1000;
  rd->thresh_mult[THR_NEARB] += 1000;
#endif  // CONFIG_EXT_REFS
  rd->thresh_mult[THR_NEARA] += 1000;
  rd->thresh_mult[THR_NEARG] += 1000;

  rd->thresh_mult[THR_ZEROMV] += 2000;
#if CONFIG_EXT_REFS
  rd->thresh_mult[THR_ZEROL2] += 2000;
  rd->thresh_mult[THR_ZEROL3] += 2000;
  rd->thresh_mult[THR_ZEROB] += 2000;
#endif  // CONFIG_EXT_REFS
  rd->thresh_mult[THR_ZEROG] += 2000;
  rd->thresh_mult[THR_ZEROA] += 2000;

  rd->thresh_mult[THR_TM] += 1000;

#if CONFIG_EXT_INTER

  rd->thresh_mult[THR_COMP_NEAREST_NEARESTLA] += 1000;
#if CONFIG_EXT_REFS
  rd->thresh_mult[THR_COMP_NEAREST_NEARESTL2A] += 1000;
  rd->thresh_mult[THR_COMP_NEAREST_NEARESTL3A] += 1000;
#endif  // CONFIG_EXT_REFS
  rd->thresh_mult[THR_COMP_NEAREST_NEARESTGA] += 1000;
#if CONFIG_EXT_REFS
  rd->thresh_mult[THR_COMP_NEAREST_NEARESTLB] += 1000;
  rd->thresh_mult[THR_COMP_NEAREST_NEARESTL2B] += 1000;
  rd->thresh_mult[THR_COMP_NEAREST_NEARESTL3B] += 1000;
  rd->thresh_mult[THR_COMP_NEAREST_NEARESTGB] += 1000;
#endif  // CONFIG_EXT_REFS

#else  // CONFIG_EXT_INTER

  rd->thresh_mult[THR_COMP_NEARESTLA] += 1000;
#if CONFIG_EXT_REFS
  rd->thresh_mult[THR_COMP_NEARESTL2A] += 1000;
  rd->thresh_mult[THR_COMP_NEARESTL3A] += 1000;
#endif  // CONFIG_EXT_REFS
  rd->thresh_mult[THR_COMP_NEARESTGA] += 1000;
#if CONFIG_EXT_REFS
  rd->thresh_mult[THR_COMP_NEARESTLB] += 1000;
  rd->thresh_mult[THR_COMP_NEARESTL2B] += 1000;
  rd->thresh_mult[THR_COMP_NEARESTL3B] += 1000;
  rd->thresh_mult[THR_COMP_NEARESTGB] += 1000;
#endif  // CONFIG_EXT_REFS

#endif  // CONFIG_EXT_INTER

#if CONFIG_EXT_INTER

  rd->thresh_mult[THR_COMP_NEAR_NEARLA] += 1200;
  rd->thresh_mult[THR_COMP_NEAREST_NEWLA] += 1500;
  rd->thresh_mult[THR_COMP_NEW_NEARESTLA] += 1500;
  rd->thresh_mult[THR_COMP_NEAR_NEWLA] += 1700;
  rd->thresh_mult[THR_COMP_NEW_NEARLA] += 1700;
  rd->thresh_mult[THR_COMP_NEW_NEWLA] += 2000;
  rd->thresh_mult[THR_COMP_ZERO_ZEROLA] += 2500;

#if CONFIG_EXT_REFS
  rd->thresh_mult[THR_COMP_NEAR_NEARL2A] += 1200;
  rd->thresh_mult[THR_COMP_NEAREST_NEWL2A] += 1500;
  rd->thresh_mult[THR_COMP_NEW_NEARESTL2A] += 1500;
  rd->thresh_mult[THR_COMP_NEAR_NEWL2A] += 1700;
  rd->thresh_mult[THR_COMP_NEW_NEARL2A] += 1700;
  rd->thresh_mult[THR_COMP_NEW_NEWL2A] += 2000;
  rd->thresh_mult[THR_COMP_ZERO_ZEROL2A] += 2500;

  rd->thresh_mult[THR_COMP_NEAR_NEARL3A] += 1200;
  rd->thresh_mult[THR_COMP_NEAREST_NEWL3A] += 1500;
  rd->thresh_mult[THR_COMP_NEW_NEARESTL3A] += 1500;
  rd->thresh_mult[THR_COMP_NEAR_NEWL3A] += 1700;
  rd->thresh_mult[THR_COMP_NEW_NEARL3A] += 1700;
  rd->thresh_mult[THR_COMP_NEW_NEWL3A] += 2000;
  rd->thresh_mult[THR_COMP_ZERO_ZEROL3A] += 2500;
#endif  // CONFIG_EXT_REFS

  rd->thresh_mult[THR_COMP_NEAR_NEARGA] += 1200;
  rd->thresh_mult[THR_COMP_NEAREST_NEWGA] += 1500;
  rd->thresh_mult[THR_COMP_NEW_NEARESTGA] += 1500;
  rd->thresh_mult[THR_COMP_NEAR_NEWGA] += 1700;
  rd->thresh_mult[THR_COMP_NEW_NEARGA] += 1700;
  rd->thresh_mult[THR_COMP_NEW_NEWGA] += 2000;
  rd->thresh_mult[THR_COMP_ZERO_ZEROGA] += 2500;

#if CONFIG_EXT_REFS
  rd->thresh_mult[THR_COMP_NEAR_NEARLB] += 1200;
  rd->thresh_mult[THR_COMP_NEAREST_NEWLB] += 1500;
  rd->thresh_mult[THR_COMP_NEW_NEARESTLB] += 1500;
  rd->thresh_mult[THR_COMP_NEAR_NEWLB] += 1700;
  rd->thresh_mult[THR_COMP_NEW_NEARLB] += 1700;
  rd->thresh_mult[THR_COMP_NEW_NEWLB] += 2000;
  rd->thresh_mult[THR_COMP_ZERO_ZEROLB] += 2500;

  rd->thresh_mult[THR_COMP_NEAR_NEARL2B] += 1200;
  rd->thresh_mult[THR_COMP_NEAREST_NEWL2B] += 1500;
  rd->thresh_mult[THR_COMP_NEW_NEARESTL2B] += 1500;
  rd->thresh_mult[THR_COMP_NEAR_NEWL2B] += 1700;
  rd->thresh_mult[THR_COMP_NEW_NEARL2B] += 1700;
  rd->thresh_mult[THR_COMP_NEW_NEWL2B] += 2000;
  rd->thresh_mult[THR_COMP_ZERO_ZEROL2B] += 2500;

  rd->thresh_mult[THR_COMP_NEAR_NEARL3B] += 1200;
  rd->thresh_mult[THR_COMP_NEAREST_NEWL3B] += 1500;
  rd->thresh_mult[THR_COMP_NEW_NEARESTL3B] += 1500;
  rd->thresh_mult[THR_COMP_NEAR_NEWL3B] += 1700;
  rd->thresh_mult[THR_COMP_NEW_NEARL3B] += 1700;
  rd->thresh_mult[THR_COMP_NEW_NEWL3B] += 2000;
  rd->thresh_mult[THR_COMP_ZERO_ZEROL3B] += 2500;

  rd->thresh_mult[THR_COMP_NEAR_NEARGB] += 1200;
  rd->thresh_mult[THR_COMP_NEAREST_NEWGB] += 1500;
  rd->thresh_mult[THR_COMP_NEW_NEARESTGB] += 1500;
  rd->thresh_mult[THR_COMP_NEAR_NEWGB] += 1700;
  rd->thresh_mult[THR_COMP_NEW_NEARGB] += 1700;
  rd->thresh_mult[THR_COMP_NEW_NEWGB] += 2000;
  rd->thresh_mult[THR_COMP_ZERO_ZEROGB] += 2500;
#endif  // CONFIG_EXT_REFS

#else  // CONFIG_EXT_INTER

  rd->thresh_mult[THR_COMP_NEARLA] += 1500;
  rd->thresh_mult[THR_COMP_NEWLA] += 2000;
#if CONFIG_EXT_REFS
  rd->thresh_mult[THR_COMP_NEARL2A] += 1500;
  rd->thresh_mult[THR_COMP_NEWL2A] += 2000;
  rd->thresh_mult[THR_COMP_NEARL3A] += 1500;
  rd->thresh_mult[THR_COMP_NEWL3A] += 2000;
#endif  // CONFIG_EXT_REFS
  rd->thresh_mult[THR_COMP_NEARGA] += 1500;
  rd->thresh_mult[THR_COMP_NEWGA] += 2000;

#if CONFIG_EXT_REFS
  rd->thresh_mult[THR_COMP_NEARLB] += 1500;
  rd->thresh_mult[THR_COMP_NEWLB] += 2000;
  rd->thresh_mult[THR_COMP_NEARL2B] += 1500;
  rd->thresh_mult[THR_COMP_NEWL2B] += 2000;
  rd->thresh_mult[THR_COMP_NEARL3B] += 1500;
  rd->thresh_mult[THR_COMP_NEWL3B] += 2000;
  rd->thresh_mult[THR_COMP_NEARGB] += 1500;
  rd->thresh_mult[THR_COMP_NEWGB] += 2000;
#endif  // CONFIG_EXT_REFS

  rd->thresh_mult[THR_COMP_ZEROLA] += 2500;
#if CONFIG_EXT_REFS
  rd->thresh_mult[THR_COMP_ZEROL2A] += 2500;
  rd->thresh_mult[THR_COMP_ZEROL3A] += 2500;
#endif  // CONFIG_EXT_REFS
  rd->thresh_mult[THR_COMP_ZEROGA] += 2500;

#if CONFIG_EXT_REFS
  rd->thresh_mult[THR_COMP_ZEROLB] += 2500;
  rd->thresh_mult[THR_COMP_ZEROL2B] += 2500;
  rd->thresh_mult[THR_COMP_ZEROL3B] += 2500;
  rd->thresh_mult[THR_COMP_ZEROGB] += 2500;
#endif  // CONFIG_EXT_REFS

#endif  // CONFIG_EXT_INTER

  rd->thresh_mult[THR_H_PRED] += 2000;
  rd->thresh_mult[THR_V_PRED] += 2000;
  rd->thresh_mult[THR_D135_PRED] += 2500;
  rd->thresh_mult[THR_D207_PRED] += 2500;
  rd->thresh_mult[THR_D153_PRED] += 2500;
  rd->thresh_mult[THR_D63_PRED] += 2500;
  rd->thresh_mult[THR_D117_PRED] += 2500;
  rd->thresh_mult[THR_D45_PRED] += 2500;

#if CONFIG_EXT_INTER
  rd->thresh_mult[THR_COMP_INTERINTRA_ZEROL] += 1500;
  rd->thresh_mult[THR_COMP_INTERINTRA_NEARESTL] += 1500;
  rd->thresh_mult[THR_COMP_INTERINTRA_NEARL] += 1500;
  rd->thresh_mult[THR_COMP_INTERINTRA_NEWL] += 2000;

#if CONFIG_EXT_REFS
  rd->thresh_mult[THR_COMP_INTERINTRA_ZEROL2] += 1500;
  rd->thresh_mult[THR_COMP_INTERINTRA_NEARESTL2] += 1500;
  rd->thresh_mult[THR_COMP_INTERINTRA_NEARL2] += 1500;
  rd->thresh_mult[THR_COMP_INTERINTRA_NEWL2] += 2000;

  rd->thresh_mult[THR_COMP_INTERINTRA_ZEROL3] += 1500;
  rd->thresh_mult[THR_COMP_INTERINTRA_NEARESTL3] += 1500;
  rd->thresh_mult[THR_COMP_INTERINTRA_NEARL3] += 1500;
  rd->thresh_mult[THR_COMP_INTERINTRA_NEWL3] += 2000;
#endif  // CONFIG_EXT_REFS

  rd->thresh_mult[THR_COMP_INTERINTRA_ZEROG] += 1500;
  rd->thresh_mult[THR_COMP_INTERINTRA_NEARESTG] += 1500;
  rd->thresh_mult[THR_COMP_INTERINTRA_NEARG] += 1500;
  rd->thresh_mult[THR_COMP_INTERINTRA_NEWG] += 2000;

#if CONFIG_EXT_REFS
  rd->thresh_mult[THR_COMP_INTERINTRA_ZEROB] += 1500;
  rd->thresh_mult[THR_COMP_INTERINTRA_NEARESTB] += 1500;
  rd->thresh_mult[THR_COMP_INTERINTRA_NEARB] += 1500;
  rd->thresh_mult[THR_COMP_INTERINTRA_NEWB] += 2000;
#endif  // CONFIG_EXT_REFS

  rd->thresh_mult[THR_COMP_INTERINTRA_ZEROA] += 1500;
  rd->thresh_mult[THR_COMP_INTERINTRA_NEARESTA] += 1500;
  rd->thresh_mult[THR_COMP_INTERINTRA_NEARA] += 1500;
  rd->thresh_mult[THR_COMP_INTERINTRA_NEWA] += 2000;
#endif  // CONFIG_EXT_INTER
}

void av1_set_rd_speed_thresholds_sub8x8(AV1_COMP *cpi) {
  static const int thresh_mult[MAX_REFS] = {
#if CONFIG_EXT_REFS
    2500,
    2500,
    2500,
    2500,
    2500,
    2500,
    4500,
    4500,
    4500,
    4500,
    4500,
    4500,
    4500,
    4500,
    2500
#else
    2500,
    2500,
    2500,
    4500,
    4500,
    2500
#endif  // CONFIG_EXT_REFS
  };
  RD_OPT *const rd = &cpi->rd;
  memcpy(rd->thresh_mult_sub8x8, thresh_mult, sizeof(thresh_mult));
}

void av1_update_rd_thresh_fact(const AV1_COMMON *const cm,
                               int (*factor_buf)[MAX_MODES], int rd_thresh,
                               int bsize, int best_mode_index) {
  if (rd_thresh > 0) {
#if CONFIG_CB4X4
    const int top_mode = MAX_MODES;
#else
    const int top_mode = bsize < BLOCK_8X8 ? MAX_REFS : MAX_MODES;
#endif
    int mode;
    for (mode = 0; mode < top_mode; ++mode) {
      const BLOCK_SIZE min_size = AOMMAX(bsize - 1, BLOCK_4X4);
      const BLOCK_SIZE max_size = AOMMIN(bsize + 2, (int)cm->sb_size);
      BLOCK_SIZE bs;
      for (bs = min_size; bs <= max_size; ++bs) {
        int *const fact = &factor_buf[bs][mode];
        if (mode == best_mode_index) {
          *fact -= (*fact >> 4);
        } else {
          *fact = AOMMIN(*fact + RD_THRESH_INC, rd_thresh * RD_THRESH_MAX_FACT);
        }
      }
    }
  }
}

int av1_get_intra_cost_penalty(int qindex, int qdelta,
                               aom_bit_depth_t bit_depth) {
  const int q = av1_dc_quant(qindex, qdelta, bit_depth);
#if CONFIG_HIGHBITDEPTH
  switch (bit_depth) {
    case AOM_BITS_8: return 20 * q;
    case AOM_BITS_10: return 5 * q;
    case AOM_BITS_12: return ROUND_POWER_OF_TWO(5 * q, 2);
    default:
      assert(0 && "bit_depth should be AOM_BITS_8, AOM_BITS_10 or AOM_BITS_12");
      return -1;
  }
#else
  return 20 * q;
#endif  // CONFIG_HIGHBITDEPTH
}