summaryrefslogtreecommitdiffstats
path: root/third_party/aom/av1/encoder/bitstream.c
blob: 2e0abc186a15b569f799208c5e522f96cb76fd54 (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
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
1958
1959
1960
1961
1962
1963
1964
1965
1966
1967
1968
1969
1970
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
2025
2026
2027
2028
2029
2030
2031
2032
2033
2034
2035
2036
2037
2038
2039
2040
2041
2042
2043
2044
2045
2046
2047
2048
2049
2050
2051
2052
2053
2054
2055
2056
2057
2058
2059
2060
2061
2062
2063
2064
2065
2066
2067
2068
2069
2070
2071
2072
2073
2074
2075
2076
2077
2078
2079
2080
2081
2082
2083
2084
2085
2086
2087
2088
2089
2090
2091
2092
2093
2094
2095
2096
2097
2098
2099
2100
2101
2102
2103
2104
2105
2106
2107
2108
2109
2110
2111
2112
2113
2114
2115
2116
2117
2118
2119
2120
2121
2122
2123
2124
2125
2126
2127
2128
2129
2130
2131
2132
2133
2134
2135
2136
2137
2138
2139
2140
2141
2142
2143
2144
2145
2146
2147
2148
2149
2150
2151
2152
2153
2154
2155
2156
2157
2158
2159
2160
2161
2162
2163
2164
2165
2166
2167
2168
2169
2170
2171
2172
2173
2174
2175
2176
2177
2178
2179
2180
2181
2182
2183
2184
2185
2186
2187
2188
2189
2190
2191
2192
2193
2194
2195
2196
2197
2198
2199
2200
2201
2202
2203
2204
2205
2206
2207
2208
2209
2210
2211
2212
2213
2214
2215
2216
2217
2218
2219
2220
2221
2222
2223
2224
2225
2226
2227
2228
2229
2230
2231
2232
2233
2234
2235
2236
2237
2238
2239
2240
2241
2242
2243
2244
2245
2246
2247
2248
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259
2260
2261
2262
2263
2264
2265
2266
2267
2268
2269
2270
2271
2272
2273
2274
2275
2276
2277
2278
2279
2280
2281
2282
2283
2284
2285
2286
2287
2288
2289
2290
2291
2292
2293
2294
2295
2296
2297
2298
2299
2300
2301
2302
2303
2304
2305
2306
2307
2308
2309
2310
2311
2312
2313
2314
2315
2316
2317
2318
2319
2320
2321
2322
2323
2324
2325
2326
2327
2328
2329
2330
2331
2332
2333
2334
2335
2336
2337
2338
2339
2340
2341
2342
2343
2344
2345
2346
2347
2348
2349
2350
2351
2352
2353
2354
2355
2356
2357
2358
2359
2360
2361
2362
2363
2364
2365
2366
2367
2368
2369
2370
2371
2372
2373
2374
2375
2376
2377
2378
2379
2380
2381
2382
2383
2384
2385
2386
2387
2388
2389
2390
2391
2392
2393
2394
2395
2396
2397
2398
2399
2400
2401
2402
2403
2404
2405
2406
2407
2408
2409
2410
2411
2412
2413
2414
2415
2416
2417
2418
2419
2420
2421
2422
2423
2424
2425
2426
2427
2428
2429
2430
2431
2432
2433
2434
2435
2436
2437
2438
2439
2440
2441
2442
2443
2444
2445
2446
2447
2448
2449
2450
2451
2452
2453
2454
2455
2456
2457
2458
2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
2475
2476
2477
2478
2479
2480
2481
2482
2483
2484
2485
2486
2487
2488
2489
2490
2491
2492
2493
2494
2495
2496
2497
2498
2499
2500
2501
2502
2503
2504
2505
2506
2507
2508
2509
2510
2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
2532
2533
2534
2535
2536
2537
2538
2539
2540
2541
2542
2543
2544
2545
2546
2547
2548
2549
2550
2551
2552
2553
2554
2555
2556
2557
2558
2559
2560
2561
2562
2563
2564
2565
2566
2567
2568
2569
2570
2571
2572
2573
2574
2575
2576
2577
2578
2579
2580
2581
2582
2583
2584
2585
2586
2587
2588
2589
2590
2591
2592
2593
2594
2595
2596
2597
2598
2599
2600
2601
2602
2603
2604
2605
2606
2607
2608
2609
2610
2611
2612
2613
2614
2615
2616
2617
2618
2619
2620
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633
2634
2635
2636
2637
2638
2639
2640
2641
2642
2643
2644
2645
2646
2647
2648
2649
2650
2651
2652
2653
2654
2655
2656
2657
2658
2659
2660
2661
2662
2663
2664
2665
2666
2667
2668
2669
2670
2671
2672
2673
2674
2675
2676
2677
2678
2679
2680
2681
2682
2683
2684
2685
2686
2687
2688
2689
2690
2691
2692
2693
2694
2695
2696
2697
2698
2699
2700
2701
2702
2703
2704
2705
2706
2707
2708
2709
2710
2711
2712
2713
2714
2715
2716
2717
2718
2719
2720
2721
2722
2723
2724
2725
2726
2727
2728
2729
2730
2731
2732
2733
2734
2735
2736
2737
2738
2739
2740
2741
2742
2743
2744
2745
2746
2747
2748
2749
2750
2751
2752
2753
2754
2755
2756
2757
2758
2759
2760
2761
2762
2763
2764
2765
2766
2767
2768
2769
2770
2771
2772
2773
2774
2775
2776
2777
2778
2779
2780
2781
2782
2783
2784
2785
2786
2787
2788
2789
2790
2791
2792
2793
2794
2795
2796
2797
2798
2799
2800
2801
2802
2803
2804
2805
2806
2807
2808
2809
2810
2811
2812
2813
2814
2815
2816
2817
2818
2819
2820
2821
2822
2823
2824
2825
2826
2827
2828
2829
2830
2831
2832
2833
2834
2835
2836
2837
2838
2839
2840
2841
2842
2843
2844
2845
2846
2847
2848
2849
2850
2851
2852
2853
2854
2855
2856
2857
2858
2859
2860
2861
2862
2863
2864
2865
2866
2867
2868
2869
2870
2871
2872
2873
2874
2875
2876
2877
2878
2879
2880
2881
2882
2883
2884
2885
2886
2887
2888
2889
2890
2891
2892
2893
2894
2895
2896
2897
2898
2899
2900
2901
2902
2903
2904
2905
2906
2907
2908
2909
2910
2911
2912
2913
2914
2915
2916
2917
2918
2919
2920
2921
2922
2923
2924
2925
2926
2927
2928
2929
2930
2931
2932
2933
2934
2935
2936
2937
2938
2939
2940
2941
2942
2943
2944
2945
2946
2947
2948
2949
2950
2951
2952
2953
2954
2955
2956
2957
2958
2959
2960
2961
2962
2963
2964
2965
2966
2967
2968
2969
2970
2971
2972
2973
2974
2975
2976
2977
2978
2979
2980
2981
2982
2983
2984
2985
2986
2987
2988
2989
2990
2991
2992
2993
2994
2995
2996
2997
2998
2999
3000
3001
3002
3003
3004
3005
3006
3007
3008
3009
3010
3011
3012
3013
3014
3015
3016
3017
3018
3019
3020
3021
3022
3023
3024
3025
3026
3027
3028
3029
3030
3031
3032
3033
3034
3035
3036
3037
3038
3039
3040
3041
3042
3043
3044
3045
3046
3047
3048
3049
3050
3051
3052
3053
3054
3055
3056
3057
3058
3059
3060
3061
3062
3063
3064
3065
3066
3067
3068
3069
3070
3071
3072
3073
3074
3075
3076
3077
3078
3079
3080
3081
3082
3083
3084
3085
3086
3087
3088
3089
3090
3091
3092
3093
3094
3095
3096
3097
3098
3099
3100
3101
3102
3103
3104
3105
3106
3107
3108
3109
3110
3111
3112
3113
3114
3115
3116
3117
3118
3119
3120
3121
3122
3123
3124
3125
3126
3127
3128
3129
3130
3131
3132
3133
3134
3135
3136
3137
3138
3139
3140
3141
3142
3143
3144
3145
3146
3147
3148
3149
3150
3151
3152
3153
3154
3155
3156
3157
3158
3159
3160
3161
3162
3163
3164
3165
3166
3167
3168
3169
3170
3171
3172
3173
3174
3175
3176
3177
3178
3179
3180
3181
3182
3183
3184
3185
3186
3187
3188
3189
3190
3191
3192
3193
3194
3195
3196
3197
3198
3199
3200
3201
3202
3203
3204
3205
3206
3207
3208
3209
3210
3211
3212
3213
3214
3215
3216
3217
3218
3219
3220
3221
3222
3223
3224
3225
3226
3227
3228
3229
3230
3231
3232
3233
3234
3235
3236
3237
3238
3239
3240
3241
3242
3243
3244
3245
3246
3247
3248
3249
3250
3251
3252
3253
3254
3255
3256
3257
3258
3259
3260
3261
3262
3263
3264
3265
3266
3267
3268
3269
3270
3271
3272
3273
3274
3275
3276
3277
3278
3279
3280
3281
3282
3283
3284
3285
3286
3287
3288
3289
3290
3291
3292
3293
3294
3295
3296
3297
3298
3299
3300
3301
3302
3303
3304
3305
3306
3307
3308
3309
3310
3311
3312
3313
3314
3315
3316
3317
3318
3319
3320
3321
3322
3323
3324
3325
3326
3327
3328
3329
3330
3331
3332
3333
3334
3335
3336
3337
3338
3339
3340
3341
3342
3343
3344
3345
3346
3347
3348
3349
3350
3351
3352
3353
3354
3355
3356
3357
3358
3359
3360
3361
3362
3363
3364
3365
3366
3367
3368
3369
3370
3371
3372
3373
3374
3375
3376
3377
3378
3379
3380
3381
3382
3383
3384
3385
3386
3387
3388
3389
3390
3391
3392
3393
3394
3395
3396
3397
3398
3399
3400
3401
3402
3403
3404
3405
3406
3407
3408
3409
3410
3411
3412
3413
3414
3415
3416
3417
3418
3419
3420
3421
3422
3423
3424
3425
3426
3427
3428
3429
3430
3431
3432
3433
3434
3435
3436
3437
3438
3439
3440
3441
3442
3443
3444
3445
3446
3447
3448
3449
3450
3451
3452
3453
3454
3455
3456
3457
3458
3459
3460
3461
3462
3463
3464
3465
3466
3467
3468
3469
3470
3471
3472
3473
3474
3475
3476
3477
3478
3479
3480
3481
3482
3483
3484
3485
3486
3487
3488
3489
3490
3491
3492
3493
3494
3495
3496
3497
3498
3499
3500
3501
3502
3503
3504
3505
3506
3507
3508
3509
3510
3511
3512
3513
3514
3515
3516
3517
3518
3519
3520
3521
3522
3523
3524
3525
3526
3527
3528
3529
3530
3531
3532
3533
3534
3535
3536
3537
3538
3539
3540
3541
3542
3543
3544
3545
3546
3547
3548
3549
3550
3551
3552
3553
3554
3555
3556
3557
3558
3559
3560
3561
3562
3563
3564
3565
3566
3567
3568
3569
3570
3571
3572
3573
3574
3575
3576
3577
3578
3579
3580
3581
3582
3583
3584
3585
3586
3587
3588
3589
3590
3591
3592
3593
3594
3595
3596
3597
3598
3599
3600
3601
3602
3603
3604
3605
3606
3607
3608
3609
3610
3611
3612
3613
3614
3615
3616
3617
3618
3619
3620
3621
3622
3623
3624
3625
3626
3627
3628
3629
3630
3631
3632
3633
3634
3635
3636
3637
3638
3639
3640
3641
3642
3643
3644
3645
3646
3647
3648
3649
3650
3651
3652
3653
3654
3655
3656
3657
3658
3659
3660
3661
3662
3663
3664
3665
3666
3667
3668
3669
3670
3671
3672
3673
3674
3675
3676
3677
3678
3679
3680
3681
3682
3683
3684
3685
3686
3687
3688
3689
3690
3691
3692
3693
3694
3695
3696
3697
3698
3699
3700
3701
3702
3703
3704
3705
3706
3707
3708
3709
3710
3711
3712
3713
3714
3715
3716
3717
3718
3719
3720
3721
3722
3723
3724
3725
3726
3727
3728
3729
3730
3731
3732
3733
3734
3735
3736
3737
3738
3739
3740
3741
3742
3743
3744
3745
3746
3747
3748
3749
3750
3751
3752
3753
3754
3755
3756
3757
3758
3759
3760
3761
3762
3763
3764
3765
3766
3767
3768
3769
3770
3771
3772
3773
3774
3775
3776
3777
3778
3779
3780
3781
3782
3783
3784
3785
3786
3787
3788
3789
3790
3791
3792
3793
3794
3795
3796
3797
3798
3799
3800
3801
3802
3803
3804
3805
3806
3807
3808
3809
3810
3811
3812
3813
3814
3815
3816
3817
3818
3819
3820
3821
3822
3823
3824
3825
3826
3827
3828
3829
3830
3831
3832
3833
3834
3835
3836
3837
3838
3839
3840
3841
3842
3843
3844
3845
3846
3847
3848
3849
3850
3851
3852
3853
3854
3855
3856
3857
3858
3859
3860
3861
3862
3863
3864
3865
3866
3867
3868
3869
3870
3871
3872
3873
3874
3875
3876
3877
3878
3879
3880
3881
3882
3883
3884
3885
3886
3887
3888
3889
3890
3891
3892
3893
3894
3895
3896
3897
3898
3899
3900
3901
3902
3903
3904
3905
3906
3907
3908
3909
3910
3911
3912
3913
3914
3915
3916
3917
3918
3919
3920
3921
3922
3923
3924
3925
3926
3927
3928
3929
3930
3931
3932
3933
3934
3935
3936
3937
3938
3939
3940
3941
3942
3943
3944
3945
3946
3947
3948
3949
3950
3951
3952
3953
3954
3955
3956
3957
3958
3959
3960
3961
3962
3963
3964
3965
3966
3967
3968
3969
3970
3971
3972
3973
3974
3975
3976
3977
3978
3979
3980
3981
3982
3983
3984
3985
3986
3987
3988
3989
3990
3991
3992
3993
3994
3995
3996
3997
3998
3999
4000
4001
4002
4003
4004
4005
4006
4007
4008
4009
4010
4011
4012
4013
4014
4015
4016
4017
4018
4019
4020
4021
4022
4023
4024
4025
4026
4027
4028
4029
4030
4031
4032
4033
4034
4035
4036
4037
4038
4039
4040
4041
4042
4043
4044
4045
4046
4047
4048
4049
4050
4051
4052
4053
4054
4055
4056
4057
4058
4059
4060
4061
4062
4063
4064
4065
4066
4067
4068
4069
4070
4071
4072
4073
4074
4075
4076
4077
4078
4079
4080
4081
4082
4083
4084
4085
4086
4087
4088
4089
4090
4091
4092
4093
4094
4095
4096
4097
4098
4099
4100
4101
4102
4103
4104
4105
4106
4107
4108
4109
4110
4111
4112
4113
4114
4115
4116
4117
4118
4119
4120
4121
4122
4123
4124
4125
4126
4127
4128
4129
4130
4131
4132
4133
4134
4135
4136
4137
4138
4139
4140
4141
4142
4143
4144
4145
4146
4147
4148
4149
4150
4151
4152
4153
4154
4155
4156
4157
4158
4159
4160
4161
4162
4163
4164
4165
4166
4167
4168
4169
4170
4171
4172
4173
4174
4175
4176
4177
4178
4179
4180
4181
4182
4183
4184
4185
4186
4187
4188
4189
4190
4191
4192
4193
4194
4195
4196
4197
4198
4199
4200
4201
4202
4203
4204
4205
4206
4207
4208
4209
4210
4211
4212
4213
4214
4215
4216
4217
4218
4219
4220
4221
4222
4223
4224
4225
4226
4227
4228
4229
4230
4231
4232
4233
4234
4235
4236
4237
4238
4239
4240
4241
4242
4243
4244
4245
4246
4247
4248
4249
4250
4251
4252
4253
4254
4255
4256
4257
4258
4259
4260
4261
4262
4263
4264
4265
4266
4267
4268
4269
4270
4271
4272
4273
4274
4275
4276
4277
4278
4279
4280
4281
4282
4283
4284
4285
4286
4287
4288
4289
4290
4291
4292
4293
4294
4295
4296
4297
4298
4299
4300
4301
4302
4303
4304
4305
4306
4307
4308
4309
4310
4311
4312
4313
4314
4315
4316
4317
4318
4319
4320
4321
4322
4323
4324
4325
4326
4327
4328
4329
4330
4331
4332
4333
4334
4335
4336
4337
4338
4339
4340
4341
4342
4343
4344
4345
4346
4347
4348
4349
4350
4351
4352
4353
4354
4355
4356
4357
4358
4359
4360
4361
4362
4363
4364
4365
4366
4367
4368
4369
4370
4371
4372
4373
4374
4375
4376
4377
4378
4379
4380
4381
4382
4383
4384
4385
4386
4387
4388
4389
4390
4391
4392
4393
4394
4395
4396
4397
4398
4399
4400
4401
4402
4403
4404
4405
4406
4407
4408
4409
4410
4411
4412
4413
4414
4415
4416
4417
4418
4419
4420
4421
4422
4423
4424
4425
4426
4427
4428
4429
4430
4431
4432
4433
4434
4435
4436
4437
4438
4439
4440
4441
4442
4443
4444
4445
4446
4447
4448
4449
4450
4451
4452
4453
4454
4455
4456
4457
4458
4459
4460
4461
4462
4463
4464
4465
4466
4467
4468
4469
4470
4471
4472
4473
4474
4475
4476
4477
4478
4479
4480
4481
4482
4483
4484
4485
4486
4487
4488
4489
4490
4491
4492
4493
4494
4495
4496
4497
4498
4499
4500
4501
4502
4503
4504
4505
4506
4507
4508
4509
4510
4511
4512
4513
4514
4515
4516
4517
4518
4519
4520
4521
4522
4523
4524
4525
4526
4527
4528
4529
4530
4531
4532
4533
4534
4535
4536
4537
4538
4539
4540
4541
4542
4543
4544
4545
4546
4547
4548
4549
4550
4551
4552
4553
4554
4555
4556
4557
4558
4559
4560
4561
4562
4563
4564
4565
4566
4567
4568
4569
4570
4571
4572
4573
4574
4575
4576
4577
4578
4579
4580
4581
4582
4583
4584
4585
4586
4587
4588
4589
4590
4591
4592
4593
4594
4595
4596
4597
4598
4599
4600
4601
4602
4603
4604
4605
4606
4607
4608
4609
4610
4611
4612
4613
4614
4615
4616
4617
4618
4619
4620
4621
4622
4623
4624
4625
4626
4627
4628
4629
4630
4631
4632
4633
4634
4635
4636
4637
4638
4639
4640
4641
4642
4643
4644
4645
4646
4647
4648
4649
4650
4651
4652
4653
4654
4655
4656
4657
4658
4659
4660
4661
4662
4663
4664
4665
4666
4667
4668
4669
4670
4671
4672
4673
4674
4675
4676
4677
4678
4679
4680
4681
4682
4683
4684
4685
4686
4687
4688
4689
4690
4691
4692
4693
4694
4695
4696
4697
4698
4699
4700
4701
4702
4703
4704
4705
4706
4707
4708
4709
4710
4711
4712
4713
4714
4715
4716
4717
4718
4719
4720
4721
4722
4723
4724
4725
4726
4727
4728
4729
4730
4731
4732
4733
4734
4735
4736
4737
4738
4739
4740
4741
4742
4743
4744
4745
4746
4747
4748
4749
4750
4751
4752
4753
4754
4755
4756
4757
4758
4759
4760
4761
4762
4763
4764
4765
4766
4767
4768
4769
4770
4771
4772
4773
4774
4775
4776
4777
4778
4779
4780
4781
4782
4783
4784
4785
4786
4787
4788
4789
4790
4791
4792
4793
4794
4795
4796
4797
4798
4799
4800
4801
4802
4803
4804
4805
4806
4807
4808
4809
4810
4811
4812
4813
4814
4815
4816
4817
4818
4819
4820
4821
4822
4823
4824
4825
4826
4827
4828
4829
4830
4831
4832
4833
4834
4835
4836
4837
4838
4839
4840
4841
4842
4843
4844
4845
4846
4847
4848
4849
4850
4851
4852
4853
4854
4855
4856
4857
4858
4859
4860
4861
4862
4863
4864
4865
4866
4867
4868
4869
4870
4871
4872
4873
4874
4875
4876
4877
4878
4879
4880
4881
4882
4883
4884
4885
4886
4887
4888
4889
4890
4891
4892
4893
4894
4895
4896
4897
4898
4899
4900
4901
4902
4903
4904
4905
4906
4907
4908
4909
4910
4911
4912
4913
4914
4915
4916
4917
4918
4919
4920
4921
4922
4923
4924
4925
4926
4927
4928
4929
4930
4931
4932
4933
4934
4935
4936
4937
4938
4939
4940
4941
4942
4943
4944
4945
4946
4947
4948
4949
4950
4951
4952
4953
4954
4955
4956
4957
4958
4959
4960
4961
4962
4963
4964
4965
4966
4967
4968
4969
4970
4971
4972
4973
4974
4975
4976
4977
4978
4979
4980
4981
4982
4983
4984
4985
4986
4987
4988
4989
/*
 * 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 <limits.h>
#include <stdio.h>

#include "aom/aom_encoder.h"
#include "aom_dsp/bitwriter_buffer.h"
#include "aom_dsp/aom_dsp_common.h"
#include "aom_dsp/binary_codes_writer.h"
#include "aom_mem/aom_mem.h"
#include "aom_ports/mem_ops.h"
#include "aom_ports/system_state.h"
#if CONFIG_BITSTREAM_DEBUG
#include "aom_util/debug_util.h"
#endif  // CONFIG_BITSTREAM_DEBUG

#if CONFIG_CDEF
#include "av1/common/cdef.h"
#endif  // CONFIG_CDEF
#include "av1/common/entropy.h"
#include "av1/common/entropymode.h"
#include "av1/common/entropymv.h"
#include "av1/common/mvref_common.h"
#include "av1/common/odintrin.h"
#include "av1/common/pred_common.h"
#include "av1/common/reconinter.h"
#if CONFIG_EXT_INTRA
#include "av1/common/reconintra.h"
#endif  // CONFIG_EXT_INTRA
#include "av1/common/seg_common.h"
#include "av1/common/tile_common.h"

#if CONFIG_ANS
#include "aom_dsp/buf_ans.h"
#endif  // CONFIG_ANS
#if CONFIG_LV_MAP
#include "av1/encoder/encodetxb.h"
#endif  // CONFIG_LV_MAP
#include "av1/encoder/bitstream.h"
#include "av1/encoder/cost.h"
#include "av1/encoder/encodemv.h"
#include "av1/encoder/mcomp.h"
#if CONFIG_PALETTE && CONFIG_PALETTE_DELTA_ENCODING
#include "av1/encoder/palette.h"
#endif  // CONFIG_PALETTE && CONFIG_PALETTE_DELTA_ENCODING
#include "av1/encoder/segmentation.h"
#include "av1/encoder/subexp.h"
#include "av1/encoder/tokenize.h"
#if CONFIG_PVQ
#include "av1/encoder/pvq_encoder.h"
#endif

#define ENC_MISMATCH_DEBUG 0

#if CONFIG_EXT_INTER && CONFIG_COMPOUND_SINGLEREF
static struct av1_token
    inter_singleref_comp_mode_encodings[INTER_SINGLEREF_COMP_MODES];
#endif  // CONFIG_EXT_INTER && CONFIG_COMPOUND_SINGLEREF

#if CONFIG_EXT_INTRA || CONFIG_FILTER_INTRA || CONFIG_PALETTE
static INLINE void write_uniform(aom_writer *w, int n, int v) {
  const int l = get_unsigned_bits(n);
  const int m = (1 << l) - n;
  if (l == 0) return;
  if (v < m) {
    aom_write_literal(w, v, l - 1);
  } else {
    aom_write_literal(w, m + ((v - m) >> 1), l - 1);
    aom_write_literal(w, (v - m) & 1, 1);
  }
}
#endif  // CONFIG_EXT_INTRA || CONFIG_FILTER_INTRA || CONFIG_PALETTE

#if CONFIG_EXT_TX
static struct av1_token ext_tx_inter_encodings[EXT_TX_SETS_INTER][TX_TYPES];
static struct av1_token ext_tx_intra_encodings[EXT_TX_SETS_INTRA][TX_TYPES];
#else
static struct av1_token ext_tx_encodings[TX_TYPES];
#endif  // CONFIG_EXT_TX
#if CONFIG_EXT_INTRA
#if CONFIG_INTRA_INTERP
static struct av1_token intra_filter_encodings[INTRA_FILTERS];
#endif  // CONFIG_INTRA_INTERP
#endif  // CONFIG_EXT_INTRA
#if CONFIG_EXT_INTER
#if CONFIG_INTERINTRA
static struct av1_token interintra_mode_encodings[INTERINTRA_MODES];
#endif
#if CONFIG_COMPOUND_SEGMENT || CONFIG_WEDGE
static struct av1_token compound_type_encodings[COMPOUND_TYPES];
#endif  // CONFIG_COMPOUND_SEGMENT || CONFIG_WEDGE
#endif  // CONFIG_EXT_INTER
#if CONFIG_MOTION_VAR || CONFIG_WARPED_MOTION
#if CONFIG_NCOBMC_ADAPT_WEIGHT
static struct av1_token ncobmc_mode_encodings[MAX_NCOBMC_MODES];
#endif
#endif  // CONFIG_MOTION_VAR || CONFIG_WARPED_MOTION
#if CONFIG_LOOP_RESTORATION
static struct av1_token switchable_restore_encodings[RESTORE_SWITCHABLE_TYPES];
#endif  // CONFIG_LOOP_RESTORATION
static void write_uncompressed_header(AV1_COMP *cpi,
                                      struct aom_write_bit_buffer *wb);
static uint32_t write_compressed_header(AV1_COMP *cpi, uint8_t *data);
static int remux_tiles(const AV1_COMMON *const cm, uint8_t *dst,
                       const uint32_t data_size, const uint32_t max_tile_size,
                       const uint32_t max_tile_col_size,
                       int *const tile_size_bytes,
                       int *const tile_col_size_bytes);

void av1_encode_token_init(void) {
#if CONFIG_EXT_TX
  int s;
#endif  // CONFIG_EXT_TX
#if CONFIG_EXT_TX
  for (s = 1; s < EXT_TX_SETS_INTER; ++s) {
    av1_tokens_from_tree(ext_tx_inter_encodings[s], av1_ext_tx_inter_tree[s]);
  }
  for (s = 1; s < EXT_TX_SETS_INTRA; ++s) {
    av1_tokens_from_tree(ext_tx_intra_encodings[s], av1_ext_tx_intra_tree[s]);
  }
#else
  av1_tokens_from_tree(ext_tx_encodings, av1_ext_tx_tree);
#endif  // CONFIG_EXT_TX

#if CONFIG_EXT_INTRA && CONFIG_INTRA_INTERP
  av1_tokens_from_tree(intra_filter_encodings, av1_intra_filter_tree);
#endif  // CONFIG_EXT_INTRA && CONFIG_INTRA_INTERP
#if CONFIG_EXT_INTER
#if CONFIG_INTERINTRA
  av1_tokens_from_tree(interintra_mode_encodings, av1_interintra_mode_tree);
#endif  // CONFIG_INTERINTRA
#if CONFIG_COMPOUND_SINGLEREF
  av1_tokens_from_tree(inter_singleref_comp_mode_encodings,
                       av1_inter_singleref_comp_mode_tree);
#endif  // CONFIG_COMPOUND_SINGLEREF
#if CONFIG_COMPOUND_SEGMENT || CONFIG_WEDGE
  av1_tokens_from_tree(compound_type_encodings, av1_compound_type_tree);
#endif  // CONFIG_COMPOUND_SEGMENT || CONFIG_WEDGE
#endif  // CONFIG_EXT_INTER
#if CONFIG_MOTION_VAR || CONFIG_WARPED_MOTION
#if CONFIG_NCOBMC_ADAPT_WEIGHT
  av1_tokens_from_tree(ncobmc_mode_encodings, av1_ncobmc_mode_tree);
#endif
#endif  // CONFIG_MOTION_VAR || CONFIG_WARPED_MOTION
#if CONFIG_LOOP_RESTORATION
  av1_tokens_from_tree(switchable_restore_encodings,
                       av1_switchable_restore_tree);
#endif  // CONFIG_LOOP_RESTORATION

  /* This hack is necessary when CONFIG_DUAL_FILTER is enabled because the five
      SWITCHABLE_FILTERS are not consecutive, e.g., 0, 1, 2, 3, 4, when doing
      an in-order traversal of the av1_switchable_interp_tree structure. */
  av1_indices_from_tree(av1_switchable_interp_ind, av1_switchable_interp_inv,
                        av1_switchable_interp_tree);
/* This hack is necessary because the four TX_TYPES are not consecutive,
    e.g., 0, 1, 2, 3, when doing an in-order traversal of the av1_ext_tx_tree
    structure. */
#if CONFIG_EXT_TX
  for (s = 1; s < EXT_TX_SETS_INTRA; ++s)
    av1_indices_from_tree(av1_ext_tx_intra_ind[s], av1_ext_tx_intra_inv[s],
                          av1_ext_tx_intra_tree[s]);
  for (s = 1; s < EXT_TX_SETS_INTER; ++s)
    av1_indices_from_tree(av1_ext_tx_inter_ind[s], av1_ext_tx_inter_inv[s],
                          av1_ext_tx_inter_tree[s]);
#else
  av1_indices_from_tree(av1_ext_tx_ind, av1_ext_tx_inv, av1_ext_tx_tree);
#endif
}

static void write_intra_mode_kf(const AV1_COMMON *cm, FRAME_CONTEXT *frame_ctx,
                                const MODE_INFO *mi, const MODE_INFO *above_mi,
                                const MODE_INFO *left_mi, int block,
                                PREDICTION_MODE mode, aom_writer *w) {
#if CONFIG_INTRABC
  assert(!is_intrabc_block(&mi->mbmi));
#endif  // CONFIG_INTRABC
  aom_write_symbol(w, av1_intra_mode_ind[mode],
                   get_y_mode_cdf(frame_ctx, mi, above_mi, left_mi, block),
                   INTRA_MODES);
  (void)cm;
}

static void write_inter_mode(aom_writer *w, PREDICTION_MODE mode,
                             FRAME_CONTEXT *ec_ctx, const int16_t mode_ctx) {
  const int16_t newmv_ctx = mode_ctx & NEWMV_CTX_MASK;

#if CONFIG_NEW_MULTISYMBOL
  aom_write_symbol(w, mode != NEWMV, ec_ctx->newmv_cdf[newmv_ctx], 2);
#else
  aom_write(w, mode != NEWMV, ec_ctx->newmv_prob[newmv_ctx]);
#endif

  if (mode != NEWMV) {
    if (mode_ctx & (1 << ALL_ZERO_FLAG_OFFSET)) {
      assert(mode == ZEROMV);
      return;
    }

    const int16_t zeromv_ctx = (mode_ctx >> ZEROMV_OFFSET) & ZEROMV_CTX_MASK;
#if CONFIG_NEW_MULTISYMBOL
    aom_write_symbol(w, mode != ZEROMV, ec_ctx->zeromv_cdf[zeromv_ctx], 2);
#else
    aom_write(w, mode != ZEROMV, ec_ctx->zeromv_prob[zeromv_ctx]);
#endif

    if (mode != ZEROMV) {
      int16_t refmv_ctx = (mode_ctx >> REFMV_OFFSET) & REFMV_CTX_MASK;

      if (mode_ctx & (1 << SKIP_NEARESTMV_OFFSET)) refmv_ctx = 6;
      if (mode_ctx & (1 << SKIP_NEARMV_OFFSET)) refmv_ctx = 7;
      if (mode_ctx & (1 << SKIP_NEARESTMV_SUB8X8_OFFSET)) refmv_ctx = 8;
#if CONFIG_NEW_MULTISYMBOL
      aom_write_symbol(w, mode != NEARESTMV, ec_ctx->refmv_cdf[refmv_ctx], 2);
#else
      aom_write(w, mode != NEARESTMV, ec_ctx->refmv_prob[refmv_ctx]);
#endif
    }
  }
}

static void write_drl_idx(FRAME_CONTEXT *ec_ctx, const MB_MODE_INFO *mbmi,
                          const MB_MODE_INFO_EXT *mbmi_ext, aom_writer *w) {
  uint8_t ref_frame_type = av1_ref_frame_type(mbmi->ref_frame);

  assert(mbmi->ref_mv_idx < 3);

#if CONFIG_EXT_INTER
#if CONFIG_COMPOUND_SINGLEREF
  if (mbmi->mode == NEWMV || mbmi->mode == NEW_NEWMV ||
      mbmi->mode == SR_NEW_NEWMV) {
#else   // !CONFIG_COMPOUND_SINGLEREF
  if (mbmi->mode == NEWMV || mbmi->mode == NEW_NEWMV) {
#endif  // CONFIG_COMPOUND_SINGLEREF
#else   // !CONFIG_EXT_INTER
  if (mbmi->mode == NEWMV) {
#endif  // CONFIG_EXT_INTER
    int idx;
    for (idx = 0; idx < 2; ++idx) {
      if (mbmi_ext->ref_mv_count[ref_frame_type] > idx + 1) {
        uint8_t drl_ctx =
            av1_drl_ctx(mbmi_ext->ref_mv_stack[ref_frame_type], idx);

#if CONFIG_NEW_MULTISYMBOL
        aom_write_symbol(w, mbmi->ref_mv_idx != idx, ec_ctx->drl_cdf[drl_ctx],
                         2);
#else
        aom_write(w, mbmi->ref_mv_idx != idx, ec_ctx->drl_prob[drl_ctx]);
#endif
        if (mbmi->ref_mv_idx == idx) return;
      }
    }
    return;
  }

  if (have_nearmv_in_inter_mode(mbmi->mode)) {
    int idx;
    // TODO(jingning): Temporary solution to compensate the NEARESTMV offset.
    for (idx = 1; idx < 3; ++idx) {
      if (mbmi_ext->ref_mv_count[ref_frame_type] > idx + 1) {
        uint8_t drl_ctx =
            av1_drl_ctx(mbmi_ext->ref_mv_stack[ref_frame_type], idx);
#if CONFIG_NEW_MULTISYMBOL
        aom_write_symbol(w, mbmi->ref_mv_idx != (idx - 1),
                         ec_ctx->drl_cdf[drl_ctx], 2);
#else
        aom_write(w, mbmi->ref_mv_idx != (idx - 1), ec_ctx->drl_prob[drl_ctx]);
#endif
        if (mbmi->ref_mv_idx == (idx - 1)) return;
      }
    }
    return;
  }
}

#if CONFIG_EXT_INTER
static void write_inter_compound_mode(AV1_COMMON *cm, MACROBLOCKD *xd,
                                      aom_writer *w, PREDICTION_MODE mode,
                                      const int16_t mode_ctx) {
  assert(is_inter_compound_mode(mode));
  (void)cm;
  aom_write_symbol(w, INTER_COMPOUND_OFFSET(mode),
                   xd->tile_ctx->inter_compound_mode_cdf[mode_ctx],
                   INTER_COMPOUND_MODES);
}

#if CONFIG_COMPOUND_SINGLEREF
static void write_inter_singleref_comp_mode(MACROBLOCKD *xd, aom_writer *w,
                                            PREDICTION_MODE mode,
                                            const int16_t mode_ctx) {
  assert(is_inter_singleref_comp_mode(mode));
  aom_cdf_prob *const inter_singleref_comp_cdf =
      xd->tile_ctx->inter_singleref_comp_mode_cdf[mode_ctx];

  aom_write_symbol(w, INTER_SINGLEREF_COMP_OFFSET(mode),
                   inter_singleref_comp_cdf, INTER_SINGLEREF_COMP_MODES);
}
#endif  // CONFIG_COMPOUND_SINGLEREF
#endif  // CONFIG_EXT_INTER

static void encode_unsigned_max(struct aom_write_bit_buffer *wb, int data,
                                int max) {
  aom_wb_write_literal(wb, data, get_unsigned_bits(max));
}

#if CONFIG_NCOBMC_ADAPT_WEIGHT
static void prob_diff_update(const aom_tree_index *tree,
                             aom_prob probs[/*n - 1*/],
                             const unsigned int counts[/* n */], int n,
                             int probwt, aom_writer *w) {
  int i;
  unsigned int branch_ct[32][2];

  // Assuming max number of probabilities <= 32
  assert(n <= 32);

  av1_tree_probs_from_distribution(tree, branch_ct, counts);
  for (i = 0; i < n - 1; ++i)
    av1_cond_prob_diff_update(w, &probs[i], branch_ct[i], probwt);
}
#endif

#if CONFIG_VAR_TX
static void write_tx_size_vartx(const AV1_COMMON *cm, MACROBLOCKD *xd,
                                const MB_MODE_INFO *mbmi, TX_SIZE tx_size,
                                int depth, int blk_row, int blk_col,
                                aom_writer *w) {
#if CONFIG_NEW_MULTISYMBOL
  FRAME_CONTEXT *ec_ctx = xd->tile_ctx;
  (void)cm;
#endif
  const int tx_row = blk_row >> 1;
  const int tx_col = blk_col >> 1;
  const int max_blocks_high = max_block_high(xd, mbmi->sb_type, 0);
  const int max_blocks_wide = max_block_wide(xd, mbmi->sb_type, 0);

  int ctx = txfm_partition_context(xd->above_txfm_context + blk_col,
                                   xd->left_txfm_context + blk_row,
                                   mbmi->sb_type, tx_size);

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

  if (depth == MAX_VARTX_DEPTH) {
    txfm_partition_update(xd->above_txfm_context + blk_col,
                          xd->left_txfm_context + blk_row, tx_size, tx_size);
    return;
  }

#if CONFIG_RECT_TX_EXT
  if (tx_size == mbmi->inter_tx_size[tx_row][tx_col] ||
      mbmi->tx_size == quarter_txsize_lookup[mbmi->sb_type]) {
#else
  if (tx_size == mbmi->inter_tx_size[tx_row][tx_col]) {
#endif
#if CONFIG_NEW_MULTISYMBOL
    aom_write_symbol(w, 0, ec_ctx->txfm_partition_cdf[ctx], 2);
#else
    aom_write(w, 0, cm->fc->txfm_partition_prob[ctx]);
#endif

    txfm_partition_update(xd->above_txfm_context + blk_col,
                          xd->left_txfm_context + blk_row, tx_size, tx_size);
    // TODO(yuec): set correct txfm partition update for qttx
  } else {
    const TX_SIZE sub_txs = sub_tx_size_map[tx_size];
    const int bsl = tx_size_wide_unit[sub_txs];
    int i;

#if CONFIG_NEW_MULTISYMBOL
    aom_write_symbol(w, 1, ec_ctx->txfm_partition_cdf[ctx], 2);
#else
    aom_write(w, 1, cm->fc->txfm_partition_prob[ctx]);
#endif

    if (tx_size == TX_8X8) {
      txfm_partition_update(xd->above_txfm_context + blk_col,
                            xd->left_txfm_context + blk_row, sub_txs, tx_size);
      return;
    }

    assert(bsl > 0);
    for (i = 0; i < 4; ++i) {
      int offsetr = blk_row + (i >> 1) * bsl;
      int offsetc = blk_col + (i & 0x01) * bsl;
      write_tx_size_vartx(cm, xd, mbmi, sub_txs, depth + 1, offsetr, offsetc,
                          w);
    }
  }
}

#if !CONFIG_NEW_MULTISYMBOL
static void update_txfm_partition_probs(AV1_COMMON *cm, aom_writer *w,
                                        FRAME_COUNTS *counts, int probwt) {
  int k;
  for (k = 0; k < TXFM_PARTITION_CONTEXTS; ++k)
    av1_cond_prob_diff_update(w, &cm->fc->txfm_partition_prob[k],
                              counts->txfm_partition[k], probwt);
}
#endif  // CONFIG_NEW_MULTISYMBOL
#endif

static void write_selected_tx_size(const AV1_COMMON *cm, const MACROBLOCKD *xd,
                                   aom_writer *w) {
  const MB_MODE_INFO *const mbmi = &xd->mi[0]->mbmi;
  const BLOCK_SIZE bsize = mbmi->sb_type;
  FRAME_CONTEXT *ec_ctx = xd->tile_ctx;
  (void)cm;
// For sub8x8 blocks the tx_size symbol does not need to be sent
#if CONFIG_CB4X4 && (CONFIG_VAR_TX || CONFIG_EXT_TX) && CONFIG_RECT_TX
  if (bsize > BLOCK_4X4) {
#else
  if (bsize >= BLOCK_8X8) {
#endif
    const TX_SIZE tx_size = mbmi->tx_size;
    const int is_inter = is_inter_block(mbmi);
    const int tx_size_ctx = get_tx_size_context(xd);
    const int tx_size_cat = is_inter ? inter_tx_size_cat_lookup[bsize]
                                     : intra_tx_size_cat_lookup[bsize];
    const TX_SIZE coded_tx_size = txsize_sqr_up_map[tx_size];
    const int depth = tx_size_to_depth(coded_tx_size);
#if CONFIG_EXT_TX && CONFIG_RECT_TX
    assert(IMPLIES(is_rect_tx(tx_size), is_rect_tx_allowed(xd, mbmi)));
#endif  // CONFIG_EXT_TX && CONFIG_RECT_TX

    aom_write_symbol(w, depth, ec_ctx->tx_size_cdf[tx_size_cat][tx_size_ctx],
                     tx_size_cat + 2);
#if CONFIG_RECT_TX_EXT && (CONFIG_EXT_TX || CONFIG_VAR_TX)
    if (is_quarter_tx_allowed(xd, mbmi, is_inter) && tx_size != coded_tx_size)
      aom_write(w, tx_size == quarter_txsize_lookup[bsize],
                cm->fc->quarter_tx_size_prob);
#endif
  }
}

#if !CONFIG_NEW_MULTISYMBOL
static void update_inter_mode_probs(AV1_COMMON *cm, aom_writer *w,
                                    FRAME_COUNTS *counts) {
  int i;
  const int probwt = cm->num_tg;
  for (i = 0; i < NEWMV_MODE_CONTEXTS; ++i)
    av1_cond_prob_diff_update(w, &cm->fc->newmv_prob[i], counts->newmv_mode[i],
                              probwt);
  for (i = 0; i < ZEROMV_MODE_CONTEXTS; ++i)
    av1_cond_prob_diff_update(w, &cm->fc->zeromv_prob[i],
                              counts->zeromv_mode[i], probwt);
  for (i = 0; i < REFMV_MODE_CONTEXTS; ++i)
    av1_cond_prob_diff_update(w, &cm->fc->refmv_prob[i], counts->refmv_mode[i],
                              probwt);
  for (i = 0; i < DRL_MODE_CONTEXTS; ++i)
    av1_cond_prob_diff_update(w, &cm->fc->drl_prob[i], counts->drl_mode[i],
                              probwt);
}
#endif

static int write_skip(const AV1_COMMON *cm, const MACROBLOCKD *xd,
                      int segment_id, const MODE_INFO *mi, aom_writer *w) {
  if (segfeature_active(&cm->seg, segment_id, SEG_LVL_SKIP)) {
    return 1;
  } else {
    const int skip = mi->mbmi.skip;
#if CONFIG_NEW_MULTISYMBOL
    FRAME_CONTEXT *ec_ctx = xd->tile_ctx;
    const int ctx = av1_get_skip_context(xd);
    aom_write_symbol(w, skip, ec_ctx->skip_cdfs[ctx], 2);
#else
    aom_write(w, skip, av1_get_skip_prob(cm, xd));
#endif
    return skip;
  }
}

static void write_is_inter(const AV1_COMMON *cm, const MACROBLOCKD *xd,
                           int segment_id, aom_writer *w, const int is_inter) {
  if (!segfeature_active(&cm->seg, segment_id, SEG_LVL_REF_FRAME)) {
#if CONFIG_NEW_MULTISYMBOL
    FRAME_CONTEXT *ec_ctx = xd->tile_ctx;
    const int ctx = av1_get_intra_inter_context(xd);
    aom_write_symbol(w, is_inter, ec_ctx->intra_inter_cdf[ctx], 2);
#else
    aom_write(w, is_inter, av1_get_intra_inter_prob(cm, xd));
#endif
  }
}

#if CONFIG_MOTION_VAR || CONFIG_WARPED_MOTION
static void write_motion_mode(const AV1_COMMON *cm, MACROBLOCKD *xd,
                              const MODE_INFO *mi, aom_writer *w) {
  const MB_MODE_INFO *mbmi = &mi->mbmi;

#if CONFIG_NCOBMC_ADAPT_WEIGHT
  MOTION_MODE last_motion_mode_allowed =
      motion_mode_allowed_wrapper(0,
#if CONFIG_GLOBAL_MOTION
                                  0, cm->global_motion,
#endif  // CONFIG_GLOBAL_MOTION
                                  mi);
#else
  MOTION_MODE last_motion_mode_allowed = motion_mode_allowed(
#if CONFIG_GLOBAL_MOTION
      0, cm->global_motion,
#endif  // CONFIG_GLOBAL_MOTION
#if CONFIG_WARPED_MOTION
      xd,
#endif
      mi);
#endif  // CONFIG_NCOBMC_ADAPT_WEIGHT
  if (last_motion_mode_allowed == SIMPLE_TRANSLATION) return;
#if CONFIG_MOTION_VAR && CONFIG_WARPED_MOTION
  if (last_motion_mode_allowed == OBMC_CAUSAL) {
#if CONFIG_NEW_MULTISYMBOL
    aom_write_symbol(w, mbmi->motion_mode == OBMC_CAUSAL,
                     xd->tile_ctx->obmc_cdf[mbmi->sb_type], 2);
#else
    aom_write(w, mbmi->motion_mode == OBMC_CAUSAL,
              cm->fc->obmc_prob[mbmi->sb_type]);
#endif
  } else {
#endif  // CONFIG_MOTION_VAR && CONFIG_WARPED_MOTION
    aom_write_symbol(w, mbmi->motion_mode,
                     xd->tile_ctx->motion_mode_cdf[mbmi->sb_type],
                     MOTION_MODES);
#if CONFIG_MOTION_VAR && CONFIG_WARPED_MOTION
  }
#endif  // CONFIG_MOTION_VAR && CONFIG_WARPED_MOTION
}

#if CONFIG_NCOBMC_ADAPT_WEIGHT
static void write_ncobmc_mode(MACROBLOCKD *xd, const MODE_INFO *mi,
                              aom_writer *w) {
  const MB_MODE_INFO *mbmi = &mi->mbmi;
  ADAPT_OVERLAP_BLOCK ao_block = adapt_overlap_block_lookup[mbmi->sb_type];
  if (mbmi->motion_mode != NCOBMC_ADAPT_WEIGHT) return;

#ifndef TRAINING_WEIGHTS
  aom_write_symbol(w, mbmi->ncobmc_mode[0],
                   xd->tile_ctx->ncobmc_mode_cdf[ao_block], MAX_NCOBMC_MODES);
  if (mi_size_wide[mbmi->sb_type] != mi_size_high[mbmi->sb_type]) {
    aom_write_symbol(w, mbmi->ncobmc_mode[1],
                     xd->tile_ctx->ncobmc_mode_cdf[ao_block], MAX_NCOBMC_MODES);
  }
#else
  int block;
  for (block = 0; block < 4; ++block)
    aom_write_symbol(w, mbmi->ncobmc_mode[0][block],
                     xd->tile_ctx->ncobmc_mode_cdf[ao_block], MAX_NCOBMC_MODES);
  if (mi_size_wide[mbmi->sb_type] != mi_size_high[mbmi->sb_type]) {
    for (block = 0; block < 4; ++block)
      aom_write_symbol(w, mbmi->ncobmc_mode[1][block],
                       xd->tile_ctx->ncobmc_mode_cdf[ao_block],
                       MAX_NCOBMC_MODES);
  }
#endif
}
#endif
#endif  // CONFIG_MOTION_VAR || CONFIG_WARPED_MOTION

#if CONFIG_DELTA_Q
static void write_delta_qindex(const AV1_COMMON *cm, const MACROBLOCKD *xd,
                               int delta_qindex, aom_writer *w) {
  int sign = delta_qindex < 0;
  int abs = sign ? -delta_qindex : delta_qindex;
  int rem_bits, thr;
  int smallval = abs < DELTA_Q_SMALL ? 1 : 0;
  FRAME_CONTEXT *ec_ctx = xd->tile_ctx;
  (void)cm;

  aom_write_symbol(w, AOMMIN(abs, DELTA_Q_SMALL), ec_ctx->delta_q_cdf,
                   DELTA_Q_PROBS + 1);

  if (!smallval) {
    rem_bits = OD_ILOG_NZ(abs - 1) - 1;
    thr = (1 << rem_bits) + 1;
    aom_write_literal(w, rem_bits, 3);
    aom_write_literal(w, abs - thr, rem_bits);
  }
  if (abs > 0) {
    aom_write_bit(w, sign);
  }
}

#if CONFIG_EXT_DELTA_Q
static void write_delta_lflevel(const AV1_COMMON *cm, const MACROBLOCKD *xd,
                                int delta_lflevel, aom_writer *w) {
  int sign = delta_lflevel < 0;
  int abs = sign ? -delta_lflevel : delta_lflevel;
  int rem_bits, thr;
  int smallval = abs < DELTA_LF_SMALL ? 1 : 0;
  FRAME_CONTEXT *ec_ctx = xd->tile_ctx;
  (void)cm;

  aom_write_symbol(w, AOMMIN(abs, DELTA_LF_SMALL), ec_ctx->delta_lf_cdf,
                   DELTA_LF_PROBS + 1);

  if (!smallval) {
    rem_bits = OD_ILOG_NZ(abs - 1) - 1;
    thr = (1 << rem_bits) + 1;
    aom_write_literal(w, rem_bits, 3);
    aom_write_literal(w, abs - thr, rem_bits);
  }
  if (abs > 0) {
    aom_write_bit(w, sign);
  }
}
#endif  // CONFIG_EXT_DELTA_Q
#endif  // CONFIG_DELTA_Q

#if !CONFIG_NEW_MULTISYMBOL
static void update_skip_probs(AV1_COMMON *cm, aom_writer *w,
                              FRAME_COUNTS *counts) {
  int k;
  const int probwt = cm->num_tg;
  for (k = 0; k < SKIP_CONTEXTS; ++k) {
    av1_cond_prob_diff_update(w, &cm->fc->skip_probs[k], counts->skip[k],
                              probwt);
  }
}
#endif

#if CONFIG_PALETTE
static void pack_palette_tokens(aom_writer *w, const TOKENEXTRA **tp, int n,
                                int num) {
  const TOKENEXTRA *p = *tp;
  write_uniform(w, n, p->token);  // The first color index.
  ++p;
  --num;
  for (int i = 0; i < num; ++i) {
    aom_write_symbol(w, p->token, p->palette_cdf, n);
    ++p;
  }
  *tp = p;
}
#endif  // CONFIG_PALETTE

#if !CONFIG_PVQ
#if CONFIG_SUPERTX
static void update_supertx_probs(AV1_COMMON *cm, int probwt, aom_writer *w) {
  const int savings_thresh = av1_cost_one(GROUP_DIFF_UPDATE_PROB) -
                             av1_cost_zero(GROUP_DIFF_UPDATE_PROB);
  int i, j;
  int savings = 0;
  int do_update = 0;
  for (i = 0; i < PARTITION_SUPERTX_CONTEXTS; ++i) {
    for (j = TX_8X8; j < TX_SIZES; ++j) {
      savings += av1_cond_prob_diff_update_savings(
          &cm->fc->supertx_prob[i][j], cm->counts.supertx[i][j], probwt);
    }
  }
  do_update = savings > savings_thresh;
  aom_write(w, do_update, GROUP_DIFF_UPDATE_PROB);
  if (do_update) {
    for (i = 0; i < PARTITION_SUPERTX_CONTEXTS; ++i) {
      for (j = TX_8X8; j < TX_SIZES; ++j) {
        av1_cond_prob_diff_update(w, &cm->fc->supertx_prob[i][j],
                                  cm->counts.supertx[i][j], probwt);
      }
    }
  }
}
#endif  // CONFIG_SUPERTX

#if CONFIG_NEW_MULTISYMBOL
static INLINE void write_coeff_extra(const aom_cdf_prob *const *cdf, int val,
                                     int n, aom_writer *w) {
  // Code the extra bits from LSB to MSB in groups of 4
  int i = 0;
  int count = 0;
  while (count < n) {
    const int size = AOMMIN(n - count, 4);
    const int mask = (1 << size) - 1;
    aom_write_cdf(w, val & mask, cdf[i++], 1 << size);
    val >>= size;
    count += size;
  }
}
#else
static INLINE void write_coeff_extra(const aom_prob *pb, int value,
                                     int num_bits, int skip_bits, aom_writer *w,
                                     TOKEN_STATS *token_stats) {
  // Code the extra bits from MSB to LSB 1 bit at a time
  int index;
  for (index = skip_bits; index < num_bits; ++index) {
    const int shift = num_bits - index - 1;
    const int bb = (value >> shift) & 1;
    aom_write_record(w, bb, pb[index], token_stats);
  }
}
#endif

#if !CONFIG_LV_MAP
static void pack_mb_tokens(aom_writer *w, const TOKENEXTRA **tp,
                           const TOKENEXTRA *const stop,
                           aom_bit_depth_t bit_depth, const TX_SIZE tx_size,
                           TOKEN_STATS *token_stats) {
  const TOKENEXTRA *p = *tp;
#if CONFIG_VAR_TX
  int count = 0;
  const int seg_eob = tx_size_2d[tx_size];
#endif

  while (p < stop && p->token != EOSB_TOKEN) {
    const int token = p->token;
    const int eob_val = p->eob_val;
    if (token == BLOCK_Z_TOKEN) {
      aom_write_symbol(w, 0, *p->head_cdf, HEAD_TOKENS + 1);
      p++;
#if CONFIG_VAR_TX
      break;
#endif
      continue;
    }

    const av1_extra_bit *const extra_bits = &av1_extra_bits[token];
    if (eob_val == LAST_EOB) {
      // Just code a flag indicating whether the value is >1 or 1.
      aom_write_bit(w, token != ONE_TOKEN);
    } else {
      int comb_symb = 2 * AOMMIN(token, TWO_TOKEN) - eob_val + p->first_val;
      aom_write_symbol(w, comb_symb, *p->head_cdf, HEAD_TOKENS + p->first_val);
    }
    if (token > ONE_TOKEN) {
      aom_write_symbol(w, token - TWO_TOKEN, *p->tail_cdf, TAIL_TOKENS);
    }

    if (extra_bits->base_val) {
      const int bit_string = p->extra;
      const int bit_string_length = extra_bits->len;  // Length of extra bits to
      const int is_cat6 = (extra_bits->base_val == CAT6_MIN_VAL);
      // be written excluding
      // the sign bit.
      int skip_bits = is_cat6
                          ? (int)sizeof(av1_cat6_prob) -
                                av1_get_cat6_extrabits_size(tx_size, bit_depth)
                          : 0;

      assert(!(bit_string >> (bit_string_length - skip_bits + 1)));
      if (bit_string_length > 0)
#if CONFIG_NEW_MULTISYMBOL
        write_coeff_extra(extra_bits->cdf, bit_string >> 1,
                          bit_string_length - skip_bits, w);
#else
        write_coeff_extra(extra_bits->prob, bit_string >> 1, bit_string_length,
                          skip_bits, w, token_stats);
#endif

      aom_write_bit_record(w, bit_string & 1, token_stats);
    }
    ++p;

#if CONFIG_VAR_TX
    ++count;
    if (eob_val == EARLY_EOB || count == seg_eob) break;
#endif
  }

  *tp = p;
}
#endif  // !CONFIG_LV_MAP
#else   // !CONFIG_PVQ
static PVQ_INFO *get_pvq_block(PVQ_QUEUE *pvq_q) {
  PVQ_INFO *pvq;

  assert(pvq_q->curr_pos <= pvq_q->last_pos);
  assert(pvq_q->curr_pos < pvq_q->buf_len);

  pvq = pvq_q->buf + pvq_q->curr_pos;
  ++pvq_q->curr_pos;

  return pvq;
}

static void pack_pvq_tokens(aom_writer *w, MACROBLOCK *const x,
                            MACROBLOCKD *const xd, int plane, BLOCK_SIZE bsize,
                            const TX_SIZE tx_size) {
  PVQ_INFO *pvq;
  int idx, idy;
  const struct macroblockd_plane *const pd = &xd->plane[plane];
  od_adapt_ctx *adapt;
  int max_blocks_wide;
  int max_blocks_high;
  int step = (1 << tx_size);

#if CONFIG_CHROMA_SUB8X8
  const BLOCK_SIZE plane_bsize =
      AOMMAX(BLOCK_4X4, get_plane_block_size(bsize, pd));
#elif CONFIG_CB4X4
  const BLOCK_SIZE plane_bsize = get_plane_block_size(bsize, pd);
#else
  const BLOCK_SIZE plane_bsize =
      get_plane_block_size(AOMMAX(BLOCK_8X8, bsize), pd);
#endif

  adapt = x->daala_enc.state.adapt;

  max_blocks_wide = max_block_wide(xd, plane_bsize, plane);
  max_blocks_high = max_block_high(xd, plane_bsize, plane);

  for (idy = 0; idy < max_blocks_high; idy += step) {
    for (idx = 0; idx < max_blocks_wide; idx += step) {
      const int is_keyframe = 0;
      const int encode_flip = 0;
      const int flip = 0;
      int i;
      const int has_dc_skip = 1;
      int *exg = &adapt->pvq.pvq_exg[plane][tx_size][0];
      int *ext = adapt->pvq.pvq_ext + tx_size * PVQ_MAX_PARTITIONS;
      generic_encoder *model = adapt->pvq.pvq_param_model;

      pvq = get_pvq_block(x->pvq_q);

      // encode block skip info
      aom_write_symbol(w, pvq->ac_dc_coded,
                       adapt->skip_cdf[2 * tx_size + (plane != 0)], 4);

      // AC coeffs coded?
      if (pvq->ac_dc_coded & AC_CODED) {
        assert(pvq->bs == tx_size);
        for (i = 0; i < pvq->nb_bands; i++) {
          if (i == 0 ||
              (!pvq->skip_rest && !(pvq->skip_dir & (1 << ((i - 1) % 3))))) {
            pvq_encode_partition(
                w, pvq->qg[i], pvq->theta[i], pvq->y + pvq->off[i],
                pvq->size[i], pvq->k[i], model, adapt, exg + i, ext + i,
                (plane != 0) * OD_TXSIZES * PVQ_MAX_PARTITIONS +
                    pvq->bs * PVQ_MAX_PARTITIONS + i,
                is_keyframe, i == 0 && (i < pvq->nb_bands - 1), pvq->skip_rest,
                encode_flip, flip);
          }
          if (i == 0 && !pvq->skip_rest && pvq->bs > 0) {
            aom_write_symbol(
                w, pvq->skip_dir,
                &adapt->pvq
                     .pvq_skip_dir_cdf[(plane != 0) + 2 * (pvq->bs - 1)][0],
                7);
          }
        }
      }
      // Encode residue of DC coeff, if exist.
      if (!has_dc_skip || (pvq->ac_dc_coded & DC_CODED)) {
        generic_encode(w, &adapt->model_dc[plane],
                       abs(pvq->dq_dc_residue) - has_dc_skip,
                       &adapt->ex_dc[plane][pvq->bs][0], 2);
      }
      if ((pvq->ac_dc_coded & DC_CODED)) {
        aom_write_bit(w, pvq->dq_dc_residue < 0);
      }
    }
  }  // for (idy = 0;
}
#endif  // !CONFIG_PVG

#if CONFIG_VAR_TX && !CONFIG_COEF_INTERLEAVE
#if CONFIG_LV_MAP
static void pack_txb_tokens(aom_writer *w,
#if CONFIG_LV_MAP
                            AV1_COMMON *cm,
#endif  // CONFIG_LV_MAP
                            const TOKENEXTRA **tp,
                            const TOKENEXTRA *const tok_end,
#if CONFIG_PVQ || CONFIG_LV_MAP
                            MACROBLOCK *const x,
#endif
                            MACROBLOCKD *xd, MB_MODE_INFO *mbmi, int plane,
                            BLOCK_SIZE plane_bsize, aom_bit_depth_t bit_depth,
                            int block, int blk_row, int blk_col,
                            TX_SIZE tx_size, TOKEN_STATS *token_stats) {
  const struct macroblockd_plane *const pd = &xd->plane[plane];
  const BLOCK_SIZE bsize = txsize_to_bsize[tx_size];
  const int tx_row = blk_row >> (1 - pd->subsampling_y);
  const int tx_col = blk_col >> (1 - pd->subsampling_x);
  TX_SIZE plane_tx_size;
  const int max_blocks_high = max_block_high(xd, plane_bsize, plane);
  const int max_blocks_wide = max_block_wide(xd, plane_bsize, plane);

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

  plane_tx_size =
      plane ? uv_txsize_lookup[bsize][mbmi->inter_tx_size[tx_row][tx_col]][0][0]
            : mbmi->inter_tx_size[tx_row][tx_col];

  if (tx_size == plane_tx_size) {
    TOKEN_STATS tmp_token_stats;
    init_token_stats(&tmp_token_stats);

#if !CONFIG_PVQ
    tran_low_t *tcoeff = BLOCK_OFFSET(x->mbmi_ext->tcoeff[plane], block);
    uint16_t eob = x->mbmi_ext->eobs[plane][block];
    TXB_CTX txb_ctx = { x->mbmi_ext->txb_skip_ctx[plane][block],
                        x->mbmi_ext->dc_sign_ctx[plane][block] };
    av1_write_coeffs_txb(cm, xd, w, blk_row, blk_col, block, plane, tx_size,
                         tcoeff, eob, &txb_ctx);
#else
    pack_pvq_tokens(w, x, xd, plane, bsize, tx_size);
#endif
#if CONFIG_RD_DEBUG
    token_stats->txb_coeff_cost_map[blk_row][blk_col] = tmp_token_stats.cost;
    token_stats->cost += tmp_token_stats.cost;
#endif
  } else {
    const TX_SIZE sub_txs = sub_tx_size_map[tx_size];
    const int bsl = tx_size_wide_unit[sub_txs];
    int i;

    assert(bsl > 0);

    for (i = 0; i < 4; ++i) {
      const int offsetr = blk_row + (i >> 1) * bsl;
      const int offsetc = blk_col + (i & 0x01) * bsl;
      const int step = tx_size_wide_unit[sub_txs] * tx_size_high_unit[sub_txs];

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

      pack_txb_tokens(w,
#if CONFIG_LV_MAP
                      cm,
#endif
                      tp, tok_end,
#if CONFIG_PVQ || CONFIG_LV_MAP
                      x,
#endif
                      xd, mbmi, plane, plane_bsize, bit_depth, block, offsetr,
                      offsetc, sub_txs, token_stats);
      block += step;
    }
  }
}
#else  // CONFIG_LV_MAP
static void pack_txb_tokens(aom_writer *w, const TOKENEXTRA **tp,
                            const TOKENEXTRA *const tok_end,
#if CONFIG_PVQ
                            MACROBLOCK *const x,
#endif
                            MACROBLOCKD *xd, MB_MODE_INFO *mbmi, int plane,
                            BLOCK_SIZE plane_bsize, aom_bit_depth_t bit_depth,
                            int block, int blk_row, int blk_col,
                            TX_SIZE tx_size, TOKEN_STATS *token_stats) {
  const struct macroblockd_plane *const pd = &xd->plane[plane];
  const BLOCK_SIZE bsize = txsize_to_bsize[tx_size];
  const int tx_row = blk_row >> (1 - pd->subsampling_y);
  const int tx_col = blk_col >> (1 - pd->subsampling_x);
  TX_SIZE plane_tx_size;
  const int max_blocks_high = max_block_high(xd, plane_bsize, plane);
  const int max_blocks_wide = max_block_wide(xd, plane_bsize, plane);

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

  plane_tx_size =
      plane ? uv_txsize_lookup[bsize][mbmi->inter_tx_size[tx_row][tx_col]][0][0]
            : mbmi->inter_tx_size[tx_row][tx_col];

  if (tx_size == plane_tx_size) {
    TOKEN_STATS tmp_token_stats;
    init_token_stats(&tmp_token_stats);
#if !CONFIG_PVQ
    pack_mb_tokens(w, tp, tok_end, bit_depth, tx_size, &tmp_token_stats);
#else
    pack_pvq_tokens(w, x, xd, plane, bsize, tx_size);
#endif
#if CONFIG_RD_DEBUG
    token_stats->txb_coeff_cost_map[blk_row][blk_col] = tmp_token_stats.cost;
    token_stats->cost += tmp_token_stats.cost;
#endif
  } else {
#if CONFIG_RECT_TX_EXT
    int is_qttx = plane_tx_size == quarter_txsize_lookup[plane_bsize];
    const TX_SIZE sub_txs = is_qttx ? plane_tx_size : sub_tx_size_map[tx_size];
#else
    const TX_SIZE sub_txs = sub_tx_size_map[tx_size];
#endif
    const int bsl = tx_size_wide_unit[sub_txs];
    int i;

    assert(bsl > 0);

    for (i = 0; i < 4; ++i) {
#if CONFIG_RECT_TX_EXT
      int is_wide_tx = tx_size_wide_unit[sub_txs] > tx_size_high_unit[sub_txs];
      const int offsetr =
          is_qttx ? (is_wide_tx ? i * tx_size_high_unit[sub_txs] : 0)
                  : blk_row + (i >> 1) * bsl;
      const int offsetc =
          is_qttx ? (is_wide_tx ? 0 : i * tx_size_wide_unit[sub_txs])
                  : blk_col + (i & 0x01) * bsl;
#else
      const int offsetr = blk_row + (i >> 1) * bsl;
      const int offsetc = blk_col + (i & 0x01) * bsl;
#endif
      const int step = tx_size_wide_unit[sub_txs] * tx_size_high_unit[sub_txs];

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

      pack_txb_tokens(w, tp, tok_end,
#if CONFIG_PVQ
                      x,
#endif
                      xd, mbmi, plane, plane_bsize, bit_depth, block, offsetr,
                      offsetc, sub_txs, token_stats);
      block += step;
    }
  }
}
#endif  // CONFIG_LV_MAP
#endif  // CONFIG_VAR_TX

static void write_segment_id(aom_writer *w, const struct segmentation *seg,
                             struct segmentation_probs *segp, int segment_id) {
  if (seg->enabled && seg->update_map) {
    aom_write_symbol(w, segment_id, segp->tree_cdf, MAX_SEGMENTS);
  }
}

#if CONFIG_NEW_MULTISYMBOL
#define WRITE_REF_BIT(bname, pname) \
  aom_write_symbol(w, bname, av1_get_pred_cdf_##pname(cm, xd), 2)
#else
#define WRITE_REF_BIT(bname, pname) \
  aom_write(w, bname, av1_get_pred_prob_##pname(cm, xd))
#endif

// This function encodes the reference frame
static void write_ref_frames(const AV1_COMMON *cm, const MACROBLOCKD *xd,
                             aom_writer *w) {
  const MB_MODE_INFO *const mbmi = &xd->mi[0]->mbmi;
  const int is_compound = has_second_ref(mbmi);
  const int segment_id = mbmi->segment_id;

  // If segment level coding of this signal is disabled...
  // or the segment allows multiple reference frame options
  if (segfeature_active(&cm->seg, segment_id, SEG_LVL_REF_FRAME)) {
    assert(!is_compound);
    assert(mbmi->ref_frame[0] ==
           get_segdata(&cm->seg, segment_id, SEG_LVL_REF_FRAME));
  } else {
    // does the feature use compound prediction or not
    // (if not specified at the frame/segment level)
    if (cm->reference_mode == REFERENCE_MODE_SELECT) {
#if !SUB8X8_COMP_REF
      if (mbmi->sb_type != BLOCK_4X4)
#endif
#if CONFIG_NEW_MULTISYMBOL
        aom_write_symbol(w, is_compound, av1_get_reference_mode_cdf(cm, xd), 2);
#else
      aom_write(w, is_compound, av1_get_reference_mode_prob(cm, xd));
#endif
    } else {
      assert((!is_compound) == (cm->reference_mode == SINGLE_REFERENCE));
    }

    if (is_compound) {
#if CONFIG_EXT_COMP_REFS
      const COMP_REFERENCE_TYPE comp_ref_type = has_uni_comp_refs(mbmi)
                                                    ? UNIDIR_COMP_REFERENCE
                                                    : BIDIR_COMP_REFERENCE;
#if USE_UNI_COMP_REFS
#if CONFIG_VAR_REFS
      if ((L_OR_L2(cm) || L3_OR_G(cm)) && BWD_OR_ALT(cm))
        if (L_AND_L2(cm) || L_AND_L3(cm) || L_AND_G(cm) || BWD_AND_ALT(cm))
#endif  // CONFIG_VAR_REFS
          aom_write(w, comp_ref_type, av1_get_comp_reference_type_prob(cm, xd));
#if CONFIG_VAR_REFS
        else
          assert(comp_ref_type == BIDIR_COMP_REFERENCE);
      else
        assert(comp_ref_type == UNIDIR_COMP_REFERENCE);
#endif  // CONFIG_VAR_REFS
#else   // !USE_UNI_COMP_REFS
      // NOTE: uni-directional comp refs disabled
      assert(comp_ref_type == BIDIR_COMP_REFERENCE);
#endif  // USE_UNI_COMP_REFS

      if (comp_ref_type == UNIDIR_COMP_REFERENCE) {
        const int bit = mbmi->ref_frame[0] == BWDREF_FRAME;
#if CONFIG_VAR_REFS
        if ((L_AND_L2(cm) || L_AND_L3(cm) || L_AND_G(cm)) && BWD_AND_ALT(cm))
#endif  // CONFIG_VAR_REFS
          aom_write(w, bit, av1_get_pred_prob_uni_comp_ref_p(cm, xd));

        if (!bit) {
          assert(mbmi->ref_frame[0] == LAST_FRAME);
#if CONFIG_VAR_REFS
          if (L_AND_L2(cm) && (L_AND_L3(cm) || L_AND_G(cm))) {
#endif  // CONFIG_VAR_REFS
            const int bit1 = mbmi->ref_frame[1] == LAST3_FRAME ||
                             mbmi->ref_frame[1] == GOLDEN_FRAME;
            aom_write(w, bit1, av1_get_pred_prob_uni_comp_ref_p1(cm, xd));

            if (bit1) {
#if CONFIG_VAR_REFS
              if (L_AND_L3(cm) && L_AND_G(cm)) {
#endif  // CONFIG_VAR_REFS
                const int bit2 = mbmi->ref_frame[1] == GOLDEN_FRAME;
                aom_write(w, bit2, av1_get_pred_prob_uni_comp_ref_p2(cm, xd));
#if CONFIG_VAR_REFS
              }
#endif  // CONFIG_VAR_REFS
            }
#if CONFIG_VAR_REFS
          }
#endif  // CONFIG_VAR_REFS
        } else {
          assert(mbmi->ref_frame[1] == ALTREF_FRAME);
        }

        return;
      }

      assert(comp_ref_type == BIDIR_COMP_REFERENCE);
#endif  // CONFIG_EXT_COMP_REFS

#if CONFIG_EXT_REFS
      const int bit = (mbmi->ref_frame[0] == GOLDEN_FRAME ||
                       mbmi->ref_frame[0] == LAST3_FRAME);
#if CONFIG_VAR_REFS
      // Test need to explicitly code (L,L2) vs (L3,G) branch node in tree
      if (L_OR_L2(cm) && L3_OR_G(cm))
#endif  // CONFIG_VAR_REFS
        WRITE_REF_BIT(bit, comp_ref_p);

      if (!bit) {
#if CONFIG_VAR_REFS
        // Test need to explicitly code (L) vs (L2) branch node in tree
        if (L_AND_L2(cm)) {
#endif  // CONFIG_VAR_REFS
          const int bit1 = mbmi->ref_frame[0] == LAST_FRAME;
          WRITE_REF_BIT(bit1, comp_ref_p1);
#if CONFIG_VAR_REFS
        }
#endif  // CONFIG_VAR_REFS
      } else {
#if CONFIG_VAR_REFS
        // Test need to explicitly code (L3) vs (G) branch node in tree
        if (L3_AND_G(cm)) {
#endif  // CONFIG_VAR_REFS
          const int bit2 = mbmi->ref_frame[0] == GOLDEN_FRAME;
          WRITE_REF_BIT(bit2, comp_ref_p2);
#if CONFIG_VAR_REFS
        }
#endif  // CONFIG_VAR_REFS
      }

#if CONFIG_VAR_REFS
      // Test need to explicitly code (BWD) vs (ALT) branch node in tree
      if (BWD_AND_ALT(cm)) {
#endif  // CONFIG_VAR_REFS
        const int bit_bwd = mbmi->ref_frame[1] == ALTREF_FRAME;
        WRITE_REF_BIT(bit_bwd, comp_bwdref_p);
#if CONFIG_VAR_REFS
      }
#endif  // CONFIG_VAR_REFS

#else   // !CONFIG_EXT_REFS
      const int bit = mbmi->ref_frame[0] == GOLDEN_FRAME;
      WRITE_REF_BIT(bit, comp_ref_p);
#endif  // CONFIG_EXT_REFS
    } else {
#if CONFIG_EXT_REFS
      const int bit0 = (mbmi->ref_frame[0] == ALTREF_FRAME ||
                        mbmi->ref_frame[0] == BWDREF_FRAME);
#if CONFIG_VAR_REFS
      // Test need to explicitly code (L,L2,L3,G) vs (BWD,ALT) branch node in
      // tree
      if ((L_OR_L2(cm) || L3_OR_G(cm)) && BWD_OR_ALT(cm))
#endif  // CONFIG_VAR_REFS
        WRITE_REF_BIT(bit0, single_ref_p1);

      if (bit0) {
#if CONFIG_VAR_REFS
        // Test need to explicitly code (BWD) vs (ALT) branch node in tree
        if (BWD_AND_ALT(cm)) {
#endif  // CONFIG_VAR_REFS
          const int bit1 = mbmi->ref_frame[0] == ALTREF_FRAME;
          WRITE_REF_BIT(bit1, single_ref_p2);
#if CONFIG_VAR_REFS
        }
#endif  // CONFIG_VAR_REFS
      } else {
        const int bit2 = (mbmi->ref_frame[0] == LAST3_FRAME ||
                          mbmi->ref_frame[0] == GOLDEN_FRAME);
#if CONFIG_VAR_REFS
        // Test need to explicitly code (L,L2) vs (L3,G) branch node in tree
        if (L_OR_L2(cm) && L3_OR_G(cm))
#endif  // CONFIG_VAR_REFS
          WRITE_REF_BIT(bit2, single_ref_p3);

        if (!bit2) {
#if CONFIG_VAR_REFS
          // Test need to explicitly code (L) vs (L2) branch node in tree
          if (L_AND_L2(cm)) {
#endif  // CONFIG_VAR_REFS
            const int bit3 = mbmi->ref_frame[0] != LAST_FRAME;
            WRITE_REF_BIT(bit3, single_ref_p4);
#if CONFIG_VAR_REFS
          }
#endif  // CONFIG_VAR_REFS
        } else {
#if CONFIG_VAR_REFS
          // Test need to explicitly code (L3) vs (G) branch node in tree
          if (L3_AND_G(cm)) {
#endif  // CONFIG_VAR_REFS
            const int bit4 = mbmi->ref_frame[0] != LAST3_FRAME;
            WRITE_REF_BIT(bit4, single_ref_p5);
#if CONFIG_VAR_REFS
          }
#endif  // CONFIG_VAR_REFS
        }
      }
#else   // !CONFIG_EXT_REFS
      const int bit0 = mbmi->ref_frame[0] != LAST_FRAME;
      WRITE_REF_BIT(bit0, single_ref_p1);

      if (bit0) {
        const int bit1 = mbmi->ref_frame[0] != GOLDEN_FRAME;
        WRITE_REF_BIT(bit1, single_ref_p2);
      }
#endif  // CONFIG_EXT_REFS
    }
  }
}

#if CONFIG_FILTER_INTRA
static void write_filter_intra_mode_info(const AV1_COMMON *const cm,
                                         const MACROBLOCKD *xd,
                                         const MB_MODE_INFO *const mbmi,
                                         int mi_row, int mi_col,
                                         aom_writer *w) {
  if (mbmi->mode == DC_PRED
#if CONFIG_PALETTE
      && mbmi->palette_mode_info.palette_size[0] == 0
#endif  // CONFIG_PALETTE
      ) {
    aom_write(w, mbmi->filter_intra_mode_info.use_filter_intra_mode[0],
              cm->fc->filter_intra_probs[0]);
    if (mbmi->filter_intra_mode_info.use_filter_intra_mode[0]) {
      const FILTER_INTRA_MODE mode =
          mbmi->filter_intra_mode_info.filter_intra_mode[0];
      write_uniform(w, FILTER_INTRA_MODES, mode);
    }
  }

#if CONFIG_CB4X4
  if (!is_chroma_reference(mi_row, mi_col, mbmi->sb_type,
                           xd->plane[1].subsampling_x,
                           xd->plane[1].subsampling_y))
    return;
#else
  (void)xd;
  (void)mi_row;
  (void)mi_col;
#endif  // CONFIG_CB4X4

  if (mbmi->uv_mode == UV_DC_PRED
#if CONFIG_PALETTE
      && mbmi->palette_mode_info.palette_size[1] == 0
#endif  // CONFIG_PALETTE
      ) {
    aom_write(w, mbmi->filter_intra_mode_info.use_filter_intra_mode[1],
              cm->fc->filter_intra_probs[1]);
    if (mbmi->filter_intra_mode_info.use_filter_intra_mode[1]) {
      const FILTER_INTRA_MODE mode =
          mbmi->filter_intra_mode_info.filter_intra_mode[1];
      write_uniform(w, FILTER_INTRA_MODES, mode);
    }
  }
}
#endif  // CONFIG_FILTER_INTRA

#if CONFIG_EXT_INTRA
static void write_intra_angle_info(const MACROBLOCKD *xd,
                                   FRAME_CONTEXT *const ec_ctx, aom_writer *w) {
  const MB_MODE_INFO *const mbmi = &xd->mi[0]->mbmi;
  const BLOCK_SIZE bsize = mbmi->sb_type;
#if CONFIG_INTRA_INTERP
  const int intra_filter_ctx = av1_get_pred_context_intra_interp(xd);
  int p_angle;
#endif  // CONFIG_INTRA_INTERP

  (void)ec_ctx;
  if (!av1_use_angle_delta(bsize)) return;

  if (av1_is_directional_mode(mbmi->mode, bsize)) {
    write_uniform(w, 2 * MAX_ANGLE_DELTA + 1,
                  MAX_ANGLE_DELTA + mbmi->angle_delta[0]);
#if CONFIG_INTRA_INTERP
    p_angle = mode_to_angle_map[mbmi->mode] + mbmi->angle_delta[0] * ANGLE_STEP;
    if (av1_is_intra_filter_switchable(p_angle)) {
      aom_write_symbol(w, mbmi->intra_filter,
                       ec_ctx->intra_filter_cdf[intra_filter_ctx],
                       INTRA_FILTERS);
    }
#endif  // CONFIG_INTRA_INTERP
  }

  if (av1_is_directional_mode(get_uv_mode(mbmi->uv_mode), bsize)) {
    write_uniform(w, 2 * MAX_ANGLE_DELTA + 1,
                  MAX_ANGLE_DELTA + mbmi->angle_delta[1]);
  }
}
#endif  // CONFIG_EXT_INTRA

static void write_mb_interp_filter(AV1_COMP *cpi, const MACROBLOCKD *xd,
                                   aom_writer *w) {
  AV1_COMMON *const cm = &cpi->common;
  const MB_MODE_INFO *const mbmi = &xd->mi[0]->mbmi;
  FRAME_CONTEXT *ec_ctx = xd->tile_ctx;

  if (!av1_is_interp_needed(xd)) {
#if CONFIG_DUAL_FILTER
    for (int i = 0; i < 4; ++i)
      assert(mbmi->interp_filter[i] == (cm->interp_filter == SWITCHABLE
                                            ? EIGHTTAP_REGULAR
                                            : cm->interp_filter));
#else
    assert(mbmi->interp_filter == (cm->interp_filter == SWITCHABLE
                                       ? EIGHTTAP_REGULAR
                                       : cm->interp_filter));
#endif  // CONFIG_DUAL_FILTER
    return;
  }
  if (cm->interp_filter == SWITCHABLE) {
#if CONFIG_DUAL_FILTER
    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);
        aom_write_symbol(w, av1_switchable_interp_ind[mbmi->interp_filter[dir]],
                         ec_ctx->switchable_interp_cdf[ctx],
                         SWITCHABLE_FILTERS);
        ++cpi->interp_filter_selected[0][mbmi->interp_filter[dir]];
      } else {
        assert(mbmi->interp_filter[dir] == EIGHTTAP_REGULAR);
      }
    }
#else
    {
      const int ctx = av1_get_pred_context_switchable_interp(xd);
      aom_write_symbol(w, av1_switchable_interp_ind[mbmi->interp_filter],
                       ec_ctx->switchable_interp_cdf[ctx], SWITCHABLE_FILTERS);
      ++cpi->interp_filter_selected[0][mbmi->interp_filter];
    }
#endif  // CONFIG_DUAL_FILTER
  }
}

#if CONFIG_PALETTE
#if CONFIG_PALETTE_DELTA_ENCODING
// Transmit color values with delta encoding. Write the first value as
// literal, and the deltas between each value and the previous one. "min_val" is
// the smallest possible value of the deltas.
static void delta_encode_palette_colors(const int *colors, int num,
                                        int bit_depth, int min_val,
                                        aom_writer *w) {
  if (num <= 0) return;
  assert(colors[0] < (1 << bit_depth));
  aom_write_literal(w, colors[0], bit_depth);
  if (num == 1) return;
  int max_delta = 0;
  int deltas[PALETTE_MAX_SIZE];
  memset(deltas, 0, sizeof(deltas));
  for (int i = 1; i < num; ++i) {
    assert(colors[i] < (1 << bit_depth));
    const int delta = colors[i] - colors[i - 1];
    deltas[i - 1] = delta;
    assert(delta >= min_val);
    if (delta > max_delta) max_delta = delta;
  }
  const int min_bits = bit_depth - 3;
  int bits = AOMMAX(av1_ceil_log2(max_delta + 1 - min_val), min_bits);
  assert(bits <= bit_depth);
  int range = (1 << bit_depth) - colors[0] - min_val;
  aom_write_literal(w, bits - min_bits, 2);
  for (int i = 0; i < num - 1; ++i) {
    aom_write_literal(w, deltas[i] - min_val, bits);
    range -= deltas[i];
    bits = AOMMIN(bits, av1_ceil_log2(range));
  }
}

// Transmit luma palette color values. First signal if each color in the color
// cache is used. Those colors that are not in the cache are transmitted with
// delta encoding.
static void write_palette_colors_y(const MACROBLOCKD *const xd,
                                   const PALETTE_MODE_INFO *const pmi,
                                   int bit_depth, aom_writer *w) {
  const int n = pmi->palette_size[0];
  const MODE_INFO *const above_mi = xd->above_mi;
  const MODE_INFO *const left_mi = xd->left_mi;
  uint16_t color_cache[2 * PALETTE_MAX_SIZE];
  const int n_cache = av1_get_palette_cache(above_mi, left_mi, 0, color_cache);
  int out_cache_colors[PALETTE_MAX_SIZE];
  uint8_t cache_color_found[2 * PALETTE_MAX_SIZE];
  const int n_out_cache =
      av1_index_color_cache(color_cache, n_cache, pmi->palette_colors, n,
                            cache_color_found, out_cache_colors);
  int n_in_cache = 0;
  for (int i = 0; i < n_cache && n_in_cache < n; ++i) {
    const int found = cache_color_found[i];
    aom_write_bit(w, found);
    n_in_cache += found;
  }
  assert(n_in_cache + n_out_cache == n);
  delta_encode_palette_colors(out_cache_colors, n_out_cache, bit_depth, 1, w);
}

// Write chroma palette color values. U channel is handled similarly to the luma
// channel. For v channel, either use delta encoding or transmit raw values
// directly, whichever costs less.
static void write_palette_colors_uv(const MACROBLOCKD *const xd,
                                    const PALETTE_MODE_INFO *const pmi,
                                    int bit_depth, aom_writer *w) {
  const int n = pmi->palette_size[1];
  const uint16_t *colors_u = pmi->palette_colors + PALETTE_MAX_SIZE;
  const uint16_t *colors_v = pmi->palette_colors + 2 * PALETTE_MAX_SIZE;
  // U channel colors.
  const MODE_INFO *const above_mi = xd->above_mi;
  const MODE_INFO *const left_mi = xd->left_mi;
  uint16_t color_cache[2 * PALETTE_MAX_SIZE];
  const int n_cache = av1_get_palette_cache(above_mi, left_mi, 1, color_cache);
  int out_cache_colors[PALETTE_MAX_SIZE];
  uint8_t cache_color_found[2 * PALETTE_MAX_SIZE];
  const int n_out_cache = av1_index_color_cache(
      color_cache, n_cache, colors_u, n, cache_color_found, out_cache_colors);
  int n_in_cache = 0;
  for (int i = 0; i < n_cache && n_in_cache < n; ++i) {
    const int found = cache_color_found[i];
    aom_write_bit(w, found);
    n_in_cache += found;
  }
  delta_encode_palette_colors(out_cache_colors, n_out_cache, bit_depth, 0, w);

  // V channel colors. Don't use color cache as the colors are not sorted.
  const int max_val = 1 << bit_depth;
  int zero_count = 0, min_bits_v = 0;
  int bits_v =
      av1_get_palette_delta_bits_v(pmi, bit_depth, &zero_count, &min_bits_v);
  const int rate_using_delta =
      2 + bit_depth + (bits_v + 1) * (n - 1) - zero_count;
  const int rate_using_raw = bit_depth * n;
  if (rate_using_delta < rate_using_raw) {  // delta encoding
    assert(colors_v[0] < (1 << bit_depth));
    aom_write_bit(w, 1);
    aom_write_literal(w, bits_v - min_bits_v, 2);
    aom_write_literal(w, colors_v[0], bit_depth);
    for (int i = 1; i < n; ++i) {
      assert(colors_v[i] < (1 << bit_depth));
      if (colors_v[i] == colors_v[i - 1]) {  // No need to signal sign bit.
        aom_write_literal(w, 0, bits_v);
        continue;
      }
      const int delta = abs((int)colors_v[i] - colors_v[i - 1]);
      const int sign_bit = colors_v[i] < colors_v[i - 1];
      if (delta <= max_val - delta) {
        aom_write_literal(w, delta, bits_v);
        aom_write_bit(w, sign_bit);
      } else {
        aom_write_literal(w, max_val - delta, bits_v);
        aom_write_bit(w, !sign_bit);
      }
    }
  } else {  // Transmit raw values.
    aom_write_bit(w, 0);
    for (int i = 0; i < n; ++i) {
      assert(colors_v[i] < (1 << bit_depth));
      aom_write_literal(w, colors_v[i], bit_depth);
    }
  }
}
#endif  // CONFIG_PALETTE_DELTA_ENCODING

static void write_palette_mode_info(const AV1_COMMON *cm, const MACROBLOCKD *xd,
                                    const MODE_INFO *const mi, aom_writer *w) {
  const MB_MODE_INFO *const mbmi = &mi->mbmi;
  const MODE_INFO *const above_mi = xd->above_mi;
  const MODE_INFO *const left_mi = xd->left_mi;
  const BLOCK_SIZE bsize = mbmi->sb_type;
  const PALETTE_MODE_INFO *const pmi = &mbmi->palette_mode_info;

  if (mbmi->mode == DC_PRED) {
    const int n = pmi->palette_size[0];
    int palette_y_mode_ctx = 0;
    if (above_mi) {
      palette_y_mode_ctx +=
          (above_mi->mbmi.palette_mode_info.palette_size[0] > 0);
    }
    if (left_mi) {
      palette_y_mode_ctx +=
          (left_mi->mbmi.palette_mode_info.palette_size[0] > 0);
    }
    aom_write(
        w, n > 0,
        av1_default_palette_y_mode_prob[bsize - BLOCK_8X8][palette_y_mode_ctx]);
    if (n > 0) {
      aom_write_symbol(w, n - PALETTE_MIN_SIZE,
                       xd->tile_ctx->palette_y_size_cdf[bsize - BLOCK_8X8],
                       PALETTE_SIZES);
#if CONFIG_PALETTE_DELTA_ENCODING
      write_palette_colors_y(xd, pmi, cm->bit_depth, w);
#else
      for (int i = 0; i < n; ++i) {
        assert(pmi->palette_colors[i] < (1 << cm->bit_depth));
        aom_write_literal(w, pmi->palette_colors[i], cm->bit_depth);
      }
#endif  // CONFIG_PALETTE_DELTA_ENCODING
    }
  }

  if (mbmi->uv_mode == UV_DC_PRED) {
    const int n = pmi->palette_size[1];
    const int palette_uv_mode_ctx = (pmi->palette_size[0] > 0);
    aom_write(w, n > 0, av1_default_palette_uv_mode_prob[palette_uv_mode_ctx]);
    if (n > 0) {
      aom_write_symbol(w, n - PALETTE_MIN_SIZE,
                       xd->tile_ctx->palette_uv_size_cdf[bsize - BLOCK_8X8],
                       PALETTE_SIZES);
#if CONFIG_PALETTE_DELTA_ENCODING
      write_palette_colors_uv(xd, pmi, cm->bit_depth, w);
#else
      for (int i = 0; i < n; ++i) {
        assert(pmi->palette_colors[PALETTE_MAX_SIZE + i] <
               (1 << cm->bit_depth));
        assert(pmi->palette_colors[2 * PALETTE_MAX_SIZE + i] <
               (1 << cm->bit_depth));
        aom_write_literal(w, pmi->palette_colors[PALETTE_MAX_SIZE + i],
                          cm->bit_depth);
        aom_write_literal(w, pmi->palette_colors[2 * PALETTE_MAX_SIZE + i],
                          cm->bit_depth);
      }
#endif  // CONFIG_PALETTE_DELTA_ENCODING
    }
  }
}
#endif  // CONFIG_PALETTE

void av1_write_tx_type(const AV1_COMMON *const cm, const MACROBLOCKD *xd,
#if CONFIG_SUPERTX
                       const int supertx_enabled,
#endif
#if CONFIG_TXK_SEL
                       int blk_row, int blk_col, int block, int plane,
                       TX_SIZE tx_size,
#endif
                       aom_writer *w) {
  MB_MODE_INFO *mbmi = &xd->mi[0]->mbmi;
  const int is_inter = is_inter_block(mbmi);
#if !CONFIG_TXK_SEL
#if CONFIG_VAR_TX
  const TX_SIZE tx_size = is_inter ? mbmi->min_tx_size : mbmi->tx_size;
#else
  const TX_SIZE tx_size = mbmi->tx_size;
#endif  // CONFIG_VAR_TX
#endif  // !CONFIG_TXK_SEL
  FRAME_CONTEXT *ec_ctx = xd->tile_ctx;

#if !CONFIG_TXK_SEL
  TX_TYPE tx_type = mbmi->tx_type;
#else
  // Only y plane's tx_type is transmitted
  if (plane > 0) return;
  PLANE_TYPE plane_type = get_plane_type(plane);
  TX_TYPE tx_type =
      av1_get_tx_type(plane_type, xd, blk_row, blk_col, block, tx_size);
#endif

  if (!FIXED_TX_TYPE) {
#if CONFIG_EXT_TX
    const TX_SIZE square_tx_size = txsize_sqr_map[tx_size];
    const BLOCK_SIZE bsize = mbmi->sb_type;
    if (get_ext_tx_types(tx_size, bsize, is_inter, cm->reduced_tx_set_used) >
            1 &&
        ((!cm->seg.enabled && cm->base_qindex > 0) ||
         (cm->seg.enabled && xd->qindex[mbmi->segment_id] > 0)) &&
        !mbmi->skip &&
#if CONFIG_SUPERTX
        !supertx_enabled &&
#endif  // CONFIG_SUPERTX
        !segfeature_active(&cm->seg, mbmi->segment_id, SEG_LVL_SKIP)) {
      const int eset =
          get_ext_tx_set(tx_size, bsize, is_inter, cm->reduced_tx_set_used);
      // eset == 0 should correspond to a set with only DCT_DCT and there
      // is no need to send the tx_type
      assert(eset > 0);
      if (is_inter) {
        assert(ext_tx_used_inter[eset][tx_type]);
        aom_write_symbol(w, av1_ext_tx_inter_ind[eset][tx_type],
                         ec_ctx->inter_ext_tx_cdf[eset][square_tx_size],
                         ext_tx_cnt_inter[eset]);
      } else if (ALLOW_INTRA_EXT_TX) {
        assert(ext_tx_used_intra[eset][tx_type]);
        aom_write_symbol(
            w, av1_ext_tx_intra_ind[eset][tx_type],
            ec_ctx->intra_ext_tx_cdf[eset][square_tx_size][mbmi->mode],
            ext_tx_cnt_intra[eset]);
      }
    }
#else
    if (tx_size < TX_32X32 &&
        ((!cm->seg.enabled && cm->base_qindex > 0) ||
         (cm->seg.enabled && xd->qindex[mbmi->segment_id] > 0)) &&
        !mbmi->skip &&
#if CONFIG_SUPERTX
        !supertx_enabled &&
#endif  // CONFIG_SUPERTX
        !segfeature_active(&cm->seg, mbmi->segment_id, SEG_LVL_SKIP)) {
      if (is_inter) {
        aom_write_symbol(w, av1_ext_tx_ind[tx_type],
                         ec_ctx->inter_ext_tx_cdf[tx_size], TX_TYPES);
      } else {
        aom_write_symbol(
            w, av1_ext_tx_ind[tx_type],
            ec_ctx->intra_ext_tx_cdf[tx_size]
                                    [intra_mode_to_tx_type_context[mbmi->mode]],
            TX_TYPES);
      }
    }
#endif  // CONFIG_EXT_TX
  }
}

static void write_intra_mode(FRAME_CONTEXT *frame_ctx, BLOCK_SIZE bsize,
                             PREDICTION_MODE mode, aom_writer *w) {
  aom_write_symbol(w, av1_intra_mode_ind[mode],
                   frame_ctx->y_mode_cdf[size_group_lookup[bsize]],
                   INTRA_MODES);
}

static void write_intra_uv_mode(FRAME_CONTEXT *frame_ctx,
                                UV_PREDICTION_MODE uv_mode,
                                PREDICTION_MODE y_mode, aom_writer *w) {
  aom_write_symbol(w, av1_intra_mode_ind[get_uv_mode(uv_mode)],
                   frame_ctx->uv_mode_cdf[y_mode], UV_INTRA_MODES);
}

#if CONFIG_CFL
static void write_cfl_alphas(FRAME_CONTEXT *const frame_ctx, int ind,
                             const CFL_SIGN_TYPE signs[CFL_SIGNS],
                             aom_writer *w) {
  // Check for uninitialized signs
  if (cfl_alpha_codes[ind][CFL_PRED_U] == 0)
    assert(signs[CFL_PRED_U] == CFL_SIGN_POS);
  if (cfl_alpha_codes[ind][CFL_PRED_V] == 0)
    assert(signs[CFL_PRED_V] == CFL_SIGN_POS);

  // Write a symbol representing a combination of alpha Cb and alpha Cr.
  aom_write_symbol(w, ind, frame_ctx->cfl_alpha_cdf, CFL_ALPHABET_SIZE);

  // Signs are only signaled for nonzero codes.
  if (cfl_alpha_codes[ind][CFL_PRED_U] != 0)
    aom_write_bit(w, signs[CFL_PRED_U]);
  if (cfl_alpha_codes[ind][CFL_PRED_V] != 0)
    aom_write_bit(w, signs[CFL_PRED_V]);
}
#endif

static void pack_inter_mode_mvs(AV1_COMP *cpi, const int mi_row,
                                const int mi_col,
#if CONFIG_SUPERTX
                                int supertx_enabled,
#endif
                                aom_writer *w) {
  AV1_COMMON *const cm = &cpi->common;
  MACROBLOCK *const x = &cpi->td.mb;
  MACROBLOCKD *const xd = &x->e_mbd;
  FRAME_CONTEXT *ec_ctx = xd->tile_ctx;
  const MODE_INFO *mi = xd->mi[0];

  const struct segmentation *const seg = &cm->seg;
  struct segmentation_probs *const segp = &ec_ctx->seg;
  const MB_MODE_INFO *const mbmi = &mi->mbmi;
  const MB_MODE_INFO_EXT *const mbmi_ext = x->mbmi_ext;
  const PREDICTION_MODE mode = mbmi->mode;
  const int segment_id = mbmi->segment_id;
  const BLOCK_SIZE bsize = mbmi->sb_type;
  const int allow_hp = cm->allow_high_precision_mv;
  const int is_inter = is_inter_block(mbmi);
  const int is_compound = has_second_ref(mbmi);
  int skip, ref;
#if CONFIG_CB4X4
  const int unify_bsize = 1;
#else
  const int unify_bsize = 0;
#endif
  (void)mi_row;
  (void)mi_col;

  if (seg->update_map) {
    if (seg->temporal_update) {
      const int pred_flag = mbmi->seg_id_predicted;
#if CONFIG_NEW_MULTISYMBOL
      aom_cdf_prob *pred_cdf = av1_get_pred_cdf_seg_id(segp, xd);
      aom_write_symbol(w, pred_flag, pred_cdf, 2);
#else
      aom_prob pred_prob = av1_get_pred_prob_seg_id(segp, xd);
      aom_write(w, pred_flag, pred_prob);
#endif
      if (!pred_flag) write_segment_id(w, seg, segp, segment_id);
    } else {
      write_segment_id(w, seg, segp, segment_id);
    }
  }

#if CONFIG_SUPERTX
  if (supertx_enabled)
    skip = mbmi->skip;
  else
    skip = write_skip(cm, xd, segment_id, mi, w);
#else
  skip = write_skip(cm, xd, segment_id, mi, w);
#endif  // CONFIG_SUPERTX
#if CONFIG_DELTA_Q
  if (cm->delta_q_present_flag) {
    int super_block_upper_left =
        ((mi_row & MAX_MIB_MASK) == 0) && ((mi_col & MAX_MIB_MASK) == 0);
    if ((bsize != BLOCK_LARGEST || skip == 0) && super_block_upper_left) {
      assert(mbmi->current_q_index > 0);
      int reduced_delta_qindex =
          (mbmi->current_q_index - xd->prev_qindex) / cm->delta_q_res;
      write_delta_qindex(cm, xd, reduced_delta_qindex, w);
      xd->prev_qindex = mbmi->current_q_index;
#if CONFIG_EXT_DELTA_Q
      if (cm->delta_lf_present_flag) {
        int reduced_delta_lflevel =
            (mbmi->current_delta_lf_from_base - xd->prev_delta_lf_from_base) /
            cm->delta_lf_res;
        write_delta_lflevel(cm, xd, reduced_delta_lflevel, w);
        xd->prev_delta_lf_from_base = mbmi->current_delta_lf_from_base;
      }
#endif  // CONFIG_EXT_DELTA_Q
    }
  }
#endif

#if CONFIG_SUPERTX
  if (!supertx_enabled)
#endif  // CONFIG_SUPERTX
    write_is_inter(cm, xd, mbmi->segment_id, w, is_inter);

  if (cm->tx_mode == TX_MODE_SELECT &&
#if CONFIG_CB4X4 && (CONFIG_VAR_TX || CONFIG_RECT_TX)
#if CONFIG_RECT_TX
      bsize > BLOCK_4X4 &&
#else
      (bsize >= BLOCK_8X8 || (bsize > BLOCK_4X4 && is_inter)) &&
#endif  // CONFIG_RECT_TX
#else
      bsize >= BLOCK_8X8 &&
#endif
#if CONFIG_SUPERTX
      !supertx_enabled &&
#endif  // CONFIG_SUPERTX
      !(is_inter && skip) && !xd->lossless[segment_id]) {
#if CONFIG_VAR_TX
    if (is_inter) {  // This implies skip flag is 0.
      const TX_SIZE max_tx_size = get_vartx_max_txsize(mbmi, bsize);
      const int bh = tx_size_high_unit[max_tx_size];
      const int bw = tx_size_wide_unit[max_tx_size];
      const int width = block_size_wide[bsize] >> tx_size_wide_log2[0];
      const int height = block_size_high[bsize] >> tx_size_wide_log2[0];
      int idx, idy;
      for (idy = 0; idy < height; idy += bh)
        for (idx = 0; idx < width; idx += bw)
          write_tx_size_vartx(cm, xd, mbmi, max_tx_size, height != width, idy,
                              idx, w);
#if CONFIG_RECT_TX_EXT
      if (is_quarter_tx_allowed(xd, mbmi, is_inter_block(mbmi)) &&
          quarter_txsize_lookup[bsize] != max_tx_size &&
          (mbmi->tx_size == quarter_txsize_lookup[bsize] ||
           mbmi->tx_size == max_tx_size)) {
        aom_write(w, mbmi->tx_size != max_tx_size,
                  cm->fc->quarter_tx_size_prob);
      }
#endif
    } else {
      set_txfm_ctxs(mbmi->tx_size, xd->n8_w, xd->n8_h, skip, xd);
      write_selected_tx_size(cm, xd, w);
    }
  } else {
    set_txfm_ctxs(mbmi->tx_size, xd->n8_w, xd->n8_h, skip, xd);
#else
    write_selected_tx_size(cm, xd, w);
#endif
  }

  if (!is_inter) {
    if (bsize >= BLOCK_8X8 || unify_bsize) {
      write_intra_mode(ec_ctx, bsize, mode, w);
    } else {
      int idx, idy;
      const int num_4x4_w = num_4x4_blocks_wide_lookup[bsize];
      const int num_4x4_h = num_4x4_blocks_high_lookup[bsize];
      for (idy = 0; idy < 2; idy += num_4x4_h) {
        for (idx = 0; idx < 2; idx += num_4x4_w) {
          const PREDICTION_MODE b_mode = mi->bmi[idy * 2 + idx].as_mode;
          write_intra_mode(ec_ctx, bsize, b_mode, w);
        }
      }
    }
#if CONFIG_CB4X4
    if (is_chroma_reference(mi_row, mi_col, bsize, xd->plane[1].subsampling_x,
                            xd->plane[1].subsampling_y)) {
      write_intra_uv_mode(ec_ctx, mbmi->uv_mode, mode, w);
#else  // !CONFIG_CB4X4
    write_intra_uv_mode(ec_ctx, mbmi->uv_mode, mode, w);
#endif  // CONFIG_CB4X4

#if CONFIG_CFL
      if (mbmi->uv_mode == UV_DC_PRED) {
        write_cfl_alphas(ec_ctx, mbmi->cfl_alpha_idx, mbmi->cfl_alpha_signs, w);
      }
#endif

#if CONFIG_CB4X4
    }
#endif

#if CONFIG_EXT_INTRA
    write_intra_angle_info(xd, ec_ctx, w);
#endif  // CONFIG_EXT_INTRA
#if CONFIG_PALETTE
    if (bsize >= BLOCK_8X8 && cm->allow_screen_content_tools)
      write_palette_mode_info(cm, xd, mi, w);
#endif  // CONFIG_PALETTE
#if CONFIG_FILTER_INTRA
    if (bsize >= BLOCK_8X8 || unify_bsize)
      write_filter_intra_mode_info(cm, xd, mbmi, mi_row, mi_col, w);
#endif  // CONFIG_FILTER_INTRA
  } else {
    int16_t mode_ctx;
    write_ref_frames(cm, xd, w);

#if CONFIG_EXT_INTER && CONFIG_COMPOUND_SINGLEREF
    if (!segfeature_active(seg, segment_id, SEG_LVL_REF_FRAME)) {
      // NOTE: Handle single ref comp mode
      if (!is_compound)
        aom_write(w, is_inter_singleref_comp_mode(mode),
                  av1_get_inter_mode_prob(cm, xd));
    }
#endif  // CONFIG_EXT_INTER && CONFIG_COMPOUND_SINGLEREF

#if CONFIG_EXT_INTER
#if CONFIG_COMPOUND_SINGLEREF
    if (is_compound || is_inter_singleref_comp_mode(mode))
#else   // !CONFIG_COMPOUND_SINGLEREF
    if (is_compound)
#endif  // CONFIG_COMPOUND_SINGLEREF
      mode_ctx = mbmi_ext->compound_mode_context[mbmi->ref_frame[0]];
    else
#endif  // CONFIG_EXT_INTER

      mode_ctx = av1_mode_context_analyzer(mbmi_ext->mode_context,
                                           mbmi->ref_frame, bsize, -1);

    // If segment skip is not enabled code the mode.
    if (!segfeature_active(seg, segment_id, SEG_LVL_SKIP)) {
      if (bsize >= BLOCK_8X8 || unify_bsize) {
#if CONFIG_EXT_INTER
        if (is_inter_compound_mode(mode))
          write_inter_compound_mode(cm, xd, w, mode, mode_ctx);
#if CONFIG_COMPOUND_SINGLEREF
        else if (is_inter_singleref_comp_mode(mode))
          write_inter_singleref_comp_mode(xd, w, mode, mode_ctx);
#endif  // CONFIG_COMPOUND_SINGLEREF
        else if (is_inter_singleref_mode(mode))
#endif  // CONFIG_EXT_INTER
          write_inter_mode(w, mode, ec_ctx, mode_ctx);

#if CONFIG_EXT_INTER
        if (mode == NEWMV || mode == NEW_NEWMV ||
#if CONFIG_COMPOUND_SINGLEREF
            mbmi->mode == SR_NEW_NEWMV ||
#endif  // CONFIG_COMPOUND_SINGLEREF
            have_nearmv_in_inter_mode(mode))
#else   // !CONFIG_EXT_INTER
        if (mode == NEARMV || mode == NEWMV)
#endif  // CONFIG_EXT_INTER
          write_drl_idx(ec_ctx, mbmi, mbmi_ext, w);
        else
          assert(mbmi->ref_mv_idx == 0);
      }
    }

#if !CONFIG_DUAL_FILTER && !CONFIG_WARPED_MOTION && !CONFIG_GLOBAL_MOTION
    write_mb_interp_filter(cpi, xd, w);
#endif  // !CONFIG_DUAL_FILTER && !CONFIG_WARPED_MOTION

    if (bsize < BLOCK_8X8 && !unify_bsize) {
#if CONFIG_COMPOUND_SINGLEREF
      /// NOTE: Single ref comp mode does not support sub8x8.
      assert(is_compound || !is_inter_singleref_comp_mode(mbmi->mode));
#endif  // CONFIG_COMPOUND_SINGLEREF
      const int num_4x4_w = num_4x4_blocks_wide_lookup[bsize];
      const int num_4x4_h = num_4x4_blocks_high_lookup[bsize];
      int idx, idy;
      for (idy = 0; idy < 2; idy += num_4x4_h) {
        for (idx = 0; idx < 2; idx += num_4x4_w) {
          const int j = idy * 2 + idx;
          const PREDICTION_MODE b_mode = mi->bmi[j].as_mode;
#if CONFIG_EXT_INTER
          if (!is_compound)
#endif  // CONFIG_EXT_INTER
            mode_ctx = av1_mode_context_analyzer(mbmi_ext->mode_context,
                                                 mbmi->ref_frame, bsize, j);
#if CONFIG_EXT_INTER
          if (is_inter_compound_mode(b_mode))
            write_inter_compound_mode(cm, xd, w, b_mode, mode_ctx);
          else if (is_inter_singleref_mode(b_mode))
#endif  // CONFIG_EXT_INTER
            write_inter_mode(w, b_mode, ec_ctx, mode_ctx);

#if CONFIG_EXT_INTER
          if (b_mode == NEWMV || b_mode == NEW_NEWMV) {
#else
          if (b_mode == NEWMV) {
#endif  // CONFIG_EXT_INTER
            for (ref = 0; ref < 1 + is_compound; ++ref) {
              int8_t rf_type = av1_ref_frame_type(mbmi->ref_frame);
              int nmv_ctx = av1_nmv_ctx(mbmi_ext->ref_mv_count[rf_type],
                                        mbmi_ext->ref_mv_stack[rf_type], ref,
                                        mbmi->ref_mv_idx);
              nmv_context *nmvc = &ec_ctx->nmvc[nmv_ctx];
              av1_encode_mv(cpi, w, &mi->bmi[j].as_mv[ref].as_mv,
#if CONFIG_EXT_INTER
                            &mi->bmi[j].ref_mv[ref].as_mv,
#else
                            &mi->bmi[j].pred_mv[ref].as_mv,
#endif  // CONFIG_EXT_INTER
                            nmvc, allow_hp);
            }
          }
#if CONFIG_EXT_INTER
          else if (b_mode == NEAREST_NEWMV || b_mode == NEAR_NEWMV) {
            int8_t rf_type = av1_ref_frame_type(mbmi->ref_frame);
            int nmv_ctx = av1_nmv_ctx(mbmi_ext->ref_mv_count[rf_type],
                                      mbmi_ext->ref_mv_stack[rf_type], 1,
                                      mbmi->ref_mv_idx);
            nmv_context *nmvc = &ec_ctx->nmvc[nmv_ctx];
            av1_encode_mv(cpi, w, &mi->bmi[j].as_mv[1].as_mv,
                          &mi->bmi[j].ref_mv[1].as_mv, nmvc, allow_hp);
          } else if (b_mode == NEW_NEARESTMV || b_mode == NEW_NEARMV) {
            int8_t rf_type = av1_ref_frame_type(mbmi->ref_frame);
            int nmv_ctx = av1_nmv_ctx(mbmi_ext->ref_mv_count[rf_type],
                                      mbmi_ext->ref_mv_stack[rf_type], 0,
                                      mbmi->ref_mv_idx);
            nmv_context *nmvc = &ec_ctx->nmvc[nmv_ctx];
            av1_encode_mv(cpi, w, &mi->bmi[j].as_mv[0].as_mv,
                          &mi->bmi[j].ref_mv[0].as_mv, nmvc, allow_hp);
          }
#endif  // CONFIG_EXT_INTER
        }
      }
    } else {
#if CONFIG_EXT_INTER
      if (mode == NEWMV || mode == NEW_NEWMV) {
#else
      if (mode == NEWMV) {
#endif  // CONFIG_EXT_INTER
        int_mv ref_mv;
        for (ref = 0; ref < 1 + is_compound; ++ref) {
          int8_t rf_type = av1_ref_frame_type(mbmi->ref_frame);
          int nmv_ctx = av1_nmv_ctx(mbmi_ext->ref_mv_count[rf_type],
                                    mbmi_ext->ref_mv_stack[rf_type], ref,
                                    mbmi->ref_mv_idx);
          nmv_context *nmvc = &ec_ctx->nmvc[nmv_ctx];
          ref_mv = mbmi_ext->ref_mvs[mbmi->ref_frame[ref]][0];
          av1_encode_mv(cpi, w, &mbmi->mv[ref].as_mv, &ref_mv.as_mv, nmvc,
                        allow_hp);
        }
#if CONFIG_EXT_INTER
      } else if (mode == NEAREST_NEWMV || mode == NEAR_NEWMV) {
        int8_t rf_type = av1_ref_frame_type(mbmi->ref_frame);
        int nmv_ctx =
            av1_nmv_ctx(mbmi_ext->ref_mv_count[rf_type],
                        mbmi_ext->ref_mv_stack[rf_type], 1, mbmi->ref_mv_idx);
        nmv_context *nmvc = &ec_ctx->nmvc[nmv_ctx];
        av1_encode_mv(cpi, w, &mbmi->mv[1].as_mv,
                      &mbmi_ext->ref_mvs[mbmi->ref_frame[1]][0].as_mv, nmvc,
                      allow_hp);
      } else if (mode == NEW_NEARESTMV || mode == NEW_NEARMV) {
        int8_t rf_type = av1_ref_frame_type(mbmi->ref_frame);
        int nmv_ctx =
            av1_nmv_ctx(mbmi_ext->ref_mv_count[rf_type],
                        mbmi_ext->ref_mv_stack[rf_type], 0, mbmi->ref_mv_idx);
        nmv_context *nmvc = &ec_ctx->nmvc[nmv_ctx];
        av1_encode_mv(cpi, w, &mbmi->mv[0].as_mv,
                      &mbmi_ext->ref_mvs[mbmi->ref_frame[0]][0].as_mv, nmvc,
                      allow_hp);
#if CONFIG_COMPOUND_SINGLEREF
      } else if (  //  mode == SR_NEAREST_NEWMV ||
          mode == SR_NEAR_NEWMV || mode == SR_ZERO_NEWMV ||
          mode == SR_NEW_NEWMV) {
        int8_t rf_type = av1_ref_frame_type(mbmi->ref_frame);
        int nmv_ctx =
            av1_nmv_ctx(mbmi_ext->ref_mv_count[rf_type],
                        mbmi_ext->ref_mv_stack[rf_type], 0, mbmi->ref_mv_idx);
        nmv_context *nmvc = &ec_ctx->nmvc[nmv_ctx];
        int_mv ref_mv = mbmi_ext->ref_mvs[mbmi->ref_frame[0]][0];
        if (mode == SR_NEW_NEWMV)
          av1_encode_mv(cpi, w, &mbmi->mv[0].as_mv, &ref_mv.as_mv, nmvc,
                        allow_hp);
        av1_encode_mv(cpi, w, &mbmi->mv[1].as_mv, &ref_mv.as_mv, nmvc,
                      allow_hp);
#endif  // CONFIG_COMPOUND_SINGLEREF
#endif  // CONFIG_EXT_INTER
      }
    }

#if CONFIG_EXT_INTER && CONFIG_INTERINTRA
    if (cpi->common.reference_mode != COMPOUND_REFERENCE &&
#if CONFIG_SUPERTX
        !supertx_enabled &&
#endif  // CONFIG_SUPERTX
        cpi->common.allow_interintra_compound && is_interintra_allowed(mbmi)) {
      const int interintra = mbmi->ref_frame[1] == INTRA_FRAME;
      const int bsize_group = size_group_lookup[bsize];
#if CONFIG_NEW_MULTISYMBOL
      aom_write_symbol(w, interintra, ec_ctx->interintra_cdf[bsize_group], 2);
#else
      aom_write(w, interintra, cm->fc->interintra_prob[bsize_group]);
#endif
      if (interintra) {
        aom_write_symbol(w, mbmi->interintra_mode,
                         ec_ctx->interintra_mode_cdf[bsize_group],
                         INTERINTRA_MODES);
        if (is_interintra_wedge_used(bsize)) {
#if CONFIG_NEW_MULTISYMBOL
          aom_write_symbol(w, mbmi->use_wedge_interintra,
                           ec_ctx->wedge_interintra_cdf[bsize], 2);
#else
          aom_write(w, mbmi->use_wedge_interintra,
                    cm->fc->wedge_interintra_prob[bsize]);
#endif
          if (mbmi->use_wedge_interintra) {
            aom_write_literal(w, mbmi->interintra_wedge_index,
                              get_wedge_bits_lookup(bsize));
            assert(mbmi->interintra_wedge_sign == 0);
          }
        }
      }
    }
#endif  // CONFIG_EXT_INTER && CONFIG_INTERINTRA

#if CONFIG_MOTION_VAR || CONFIG_WARPED_MOTION
#if CONFIG_SUPERTX
    if (!supertx_enabled)
#endif  // CONFIG_SUPERTX
#if CONFIG_EXT_INTER
      if (mbmi->ref_frame[1] != INTRA_FRAME)
#endif  // CONFIG_EXT_INTER
        write_motion_mode(cm, xd, mi, w);
#if CONFIG_NCOBMC_ADAPT_WEIGHT
    write_ncobmc_mode(xd, mi, w);
#endif
#endif  // CONFIG_MOTION_VAR || CONFIG_WARPED_MOTION

#if CONFIG_EXT_INTER
    if (
#if CONFIG_COMPOUND_SINGLEREF
        is_inter_anyref_comp_mode(mbmi->mode) &&
#else   // !CONFIG_COMPOUND_SINGLEREF
        cpi->common.reference_mode != SINGLE_REFERENCE &&
        is_inter_compound_mode(mbmi->mode) &&
#endif  // CONFIG_COMPOUND_SINGLEREF
#if CONFIG_MOTION_VAR
        mbmi->motion_mode == SIMPLE_TRANSLATION &&
#endif  // CONFIG_MOTION_VAR
        is_any_masked_compound_used(bsize)) {
#if CONFIG_COMPOUND_SEGMENT || CONFIG_WEDGE
      if (cm->allow_masked_compound) {
        aom_write_symbol(w, mbmi->interinter_compound_type,
                         ec_ctx->compound_type_cdf[bsize], COMPOUND_TYPES);
#if CONFIG_WEDGE
        if (mbmi->interinter_compound_type == COMPOUND_WEDGE) {
          aom_write_literal(w, mbmi->wedge_index, get_wedge_bits_lookup(bsize));
          aom_write_bit(w, mbmi->wedge_sign);
        }
#endif  // CONFIG_WEDGE
#if CONFIG_COMPOUND_SEGMENT
        if (mbmi->interinter_compound_type == COMPOUND_SEG) {
          aom_write_literal(w, mbmi->mask_type, MAX_SEG_MASK_BITS);
        }
#endif  // CONFIG_COMPOUND_SEGMENT
      }
#endif  // CONFIG_COMPOUND_SEGMENT || CONFIG_WEDGE
    }
#endif  // CONFIG_EXT_INTER

#if CONFIG_DUAL_FILTER || CONFIG_WARPED_MOTION || CONFIG_GLOBAL_MOTION
    write_mb_interp_filter(cpi, xd, w);
#endif  // CONFIG_DUAL_FILTE || CONFIG_WARPED_MOTION
  }

#if !CONFIG_TXK_SEL
  av1_write_tx_type(cm, xd,
#if CONFIG_SUPERTX
                    supertx_enabled,
#endif
                    w);
#endif  // !CONFIG_TXK_SEL
}

static void write_mb_modes_kf(AV1_COMMON *cm,
#if CONFIG_DELTA_Q
                              MACROBLOCKD *xd,
#else
                              const MACROBLOCKD *xd,
#endif  // CONFIG_DELTA_Q
#if CONFIG_INTRABC
                              const MB_MODE_INFO_EXT *mbmi_ext,
#endif  // CONFIG_INTRABC
                              const int mi_row, const int mi_col,
                              aom_writer *w) {
  FRAME_CONTEXT *ec_ctx = xd->tile_ctx;
  const struct segmentation *const seg = &cm->seg;
  struct segmentation_probs *const segp = &ec_ctx->seg;
  const MODE_INFO *const mi = xd->mi[0];
  const MODE_INFO *const above_mi = xd->above_mi;
  const MODE_INFO *const left_mi = xd->left_mi;
  const MB_MODE_INFO *const mbmi = &mi->mbmi;
  const BLOCK_SIZE bsize = mbmi->sb_type;
#if CONFIG_CB4X4
  const int unify_bsize = 1;
#else
  const int unify_bsize = 0;
#endif
  (void)mi_row;
  (void)mi_col;

  if (seg->update_map) write_segment_id(w, seg, segp, mbmi->segment_id);

#if CONFIG_DELTA_Q
  const int skip = write_skip(cm, xd, mbmi->segment_id, mi, w);
  if (cm->delta_q_present_flag) {
    int super_block_upper_left =
        ((mi_row & MAX_MIB_MASK) == 0) && ((mi_col & MAX_MIB_MASK) == 0);
    if ((bsize != BLOCK_LARGEST || skip == 0) && super_block_upper_left) {
      assert(mbmi->current_q_index > 0);
      int reduced_delta_qindex =
          (mbmi->current_q_index - xd->prev_qindex) / cm->delta_q_res;
      write_delta_qindex(cm, xd, reduced_delta_qindex, w);
      xd->prev_qindex = mbmi->current_q_index;
#if CONFIG_EXT_DELTA_Q
      if (cm->delta_lf_present_flag) {
        int reduced_delta_lflevel =
            (mbmi->current_delta_lf_from_base - xd->prev_delta_lf_from_base) /
            cm->delta_lf_res;
        write_delta_lflevel(cm, xd, reduced_delta_lflevel, w);
        xd->prev_delta_lf_from_base = mbmi->current_delta_lf_from_base;
      }
#endif  // CONFIG_EXT_DELTA_Q
    }
  }
#else
  write_skip(cm, xd, mbmi->segment_id, mi, w);
#endif

  int enable_tx_size = cm->tx_mode == TX_MODE_SELECT &&
#if CONFIG_CB4X4 && (CONFIG_VAR_TX || CONFIG_RECT_TX)
#if CONFIG_RECT_TX
                       bsize > BLOCK_4X4 &&
#else
                       bsize >= BLOCK_8X8 &&
#endif  // CONFIG_RECT_TX
#else
                       bsize >= BLOCK_8X8 &&
#endif
                       !xd->lossless[mbmi->segment_id];

#if CONFIG_INTRABC
  if (bsize >= BLOCK_8X8 && cm->allow_screen_content_tools) {
    int use_intrabc = is_intrabc_block(mbmi);
    aom_write(w, use_intrabc, ec_ctx->intrabc_prob);
    if (use_intrabc) {
      assert(mbmi->mode == DC_PRED);
      assert(mbmi->uv_mode == UV_DC_PRED);
      if (enable_tx_size && !mbmi->skip) write_selected_tx_size(cm, xd, w);
      int_mv dv_ref = mbmi_ext->ref_mvs[INTRA_FRAME][0];
      av1_encode_dv(w, &mbmi->mv[0].as_mv, &dv_ref.as_mv, &ec_ctx->ndvc);
#if CONFIG_EXT_TX && !CONFIG_TXK_SEL
      av1_write_tx_type(cm, xd,
#if CONFIG_SUPERTX
                        0,
#endif
                        w);
#endif  // CONFIG_EXT_TX && !CONFIG_TXK_SEL
      return;
    }
  }
#endif  // CONFIG_INTRABC
  if (enable_tx_size) write_selected_tx_size(cm, xd, w);

  if (bsize >= BLOCK_8X8 || unify_bsize) {
    write_intra_mode_kf(cm, ec_ctx, mi, above_mi, left_mi, 0, mbmi->mode, w);
  } else {
    const int num_4x4_w = num_4x4_blocks_wide_lookup[bsize];
    const int num_4x4_h = num_4x4_blocks_high_lookup[bsize];
    int idx, idy;

    for (idy = 0; idy < 2; idy += num_4x4_h) {
      for (idx = 0; idx < 2; idx += num_4x4_w) {
        const int block = idy * 2 + idx;
        write_intra_mode_kf(cm, ec_ctx, mi, above_mi, left_mi, block,
                            mi->bmi[block].as_mode, w);
      }
    }
  }

#if CONFIG_CB4X4
  if (is_chroma_reference(mi_row, mi_col, bsize, xd->plane[1].subsampling_x,
                          xd->plane[1].subsampling_y)) {
    write_intra_uv_mode(ec_ctx, mbmi->uv_mode, mbmi->mode, w);
#else  // !CONFIG_CB4X4
  write_intra_uv_mode(ec_ctx, mbmi->uv_mode, mbmi->mode, w);
#endif  // CONFIG_CB4X4

#if CONFIG_CFL
    if (mbmi->uv_mode == UV_DC_PRED) {
      write_cfl_alphas(ec_ctx, mbmi->cfl_alpha_idx, mbmi->cfl_alpha_signs, w);
    }
#endif

#if CONFIG_CB4X4
  }
#endif
#if CONFIG_EXT_INTRA
  write_intra_angle_info(xd, ec_ctx, w);
#endif  // CONFIG_EXT_INTRA
#if CONFIG_PALETTE
  if (bsize >= BLOCK_8X8 && cm->allow_screen_content_tools)
    write_palette_mode_info(cm, xd, mi, w);
#endif  // CONFIG_PALETTE
#if CONFIG_FILTER_INTRA
  if (bsize >= BLOCK_8X8 || unify_bsize)
    write_filter_intra_mode_info(cm, xd, mbmi, mi_row, mi_col, w);
#endif  // CONFIG_FILTER_INTRA

#if !CONFIG_TXK_SEL
  av1_write_tx_type(cm, xd,
#if CONFIG_SUPERTX
                    0,
#endif
                    w);
#endif  // !CONFIG_TXK_SEL
}

#if CONFIG_SUPERTX
#define write_modes_b_wrapper(cpi, tile, w, tok, tok_end, supertx_enabled, \
                              mi_row, mi_col)                              \
  write_modes_b(cpi, tile, w, tok, tok_end, supertx_enabled, mi_row, mi_col)
#else
#define write_modes_b_wrapper(cpi, tile, w, tok, tok_end, supertx_enabled, \
                              mi_row, mi_col)                              \
  write_modes_b(cpi, tile, w, tok, tok_end, mi_row, mi_col)
#endif  // CONFIG_SUPERTX

#if CONFIG_RD_DEBUG
static void dump_mode_info(MODE_INFO *mi) {
  printf("\nmi->mbmi.mi_row == %d\n", mi->mbmi.mi_row);
  printf("&& mi->mbmi.mi_col == %d\n", mi->mbmi.mi_col);
  printf("&& mi->mbmi.sb_type == %d\n", mi->mbmi.sb_type);
  printf("&& mi->mbmi.tx_size == %d\n", mi->mbmi.tx_size);
  if (mi->mbmi.sb_type >= BLOCK_8X8) {
    printf("&& mi->mbmi.mode == %d\n", mi->mbmi.mode);
  } else {
    printf("&& mi->bmi[0].as_mode == %d\n", mi->bmi[0].as_mode);
  }
}
static int rd_token_stats_mismatch(RD_STATS *rd_stats, TOKEN_STATS *token_stats,
                                   int plane) {
  if (rd_stats->txb_coeff_cost[plane] != token_stats->cost) {
#if CONFIG_VAR_TX
    int r, c;
#endif
    printf("\nplane %d rd_stats->txb_coeff_cost %d token_stats->cost %d\n",
           plane, rd_stats->txb_coeff_cost[plane], token_stats->cost);
#if CONFIG_VAR_TX
    printf("rd txb_coeff_cost_map\n");
    for (r = 0; r < TXB_COEFF_COST_MAP_SIZE; ++r) {
      for (c = 0; c < TXB_COEFF_COST_MAP_SIZE; ++c) {
        printf("%d ", rd_stats->txb_coeff_cost_map[plane][r][c]);
      }
      printf("\n");
    }

    printf("pack txb_coeff_cost_map\n");
    for (r = 0; r < TXB_COEFF_COST_MAP_SIZE; ++r) {
      for (c = 0; c < TXB_COEFF_COST_MAP_SIZE; ++c) {
        printf("%d ", token_stats->txb_coeff_cost_map[r][c]);
      }
      printf("\n");
    }
#endif
    return 1;
  }
  return 0;
}
#endif

#if ENC_MISMATCH_DEBUG
static void enc_dump_logs(AV1_COMP *cpi, int mi_row, int mi_col) {
  AV1_COMMON *const cm = &cpi->common;
  MACROBLOCKD *const xd = &cpi->td.mb.e_mbd;
  MODE_INFO *m;
  xd->mi = cm->mi_grid_visible + (mi_row * cm->mi_stride + mi_col);
  m = xd->mi[0];
  if (is_inter_block(&m->mbmi)) {
#define FRAME_TO_CHECK 1
    if (cm->current_video_frame == FRAME_TO_CHECK /* && cm->show_frame == 1*/) {
      const MB_MODE_INFO *const mbmi = &m->mbmi;
      const BLOCK_SIZE bsize = mbmi->sb_type;

      int_mv mv[2];
      int is_comp_ref = has_second_ref(&m->mbmi);
      int ref;

      for (ref = 0; ref < 1 + is_comp_ref; ++ref)
        mv[ref].as_mv = m->mbmi.mv[ref].as_mv;

      if (!is_comp_ref) {
#if CONFIG_COMPOUND_SINGLEREF
        if (is_inter_singleref_comp_mode(m->mbmi.mode))
          mv[1].as_mv = m->mbmi.mv[1].as_mv;
        else
#endif  // CONFIG_COMPOUND_SINGLEREF
          mv[1].as_int = 0;
      }
      int interp_ctx[2] = { -1 };
      int interp_filter[2] = { cm->interp_filter };
      if (cm->interp_filter == SWITCHABLE) {
        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))) {
            interp_ctx[dir] = av1_get_pred_context_switchable_interp(xd, dir);
            interp_filter[dir] = mbmi->interp_filter[dir];
          } else {
            interp_filter[dir] = EIGHTTAP_REGULAR;
          }
        }
      }

      MACROBLOCK *const x = &cpi->td.mb;
      const MB_MODE_INFO_EXT *const mbmi_ext = x->mbmi_ext;
      const int16_t mode_ctx = av1_mode_context_analyzer(
          mbmi_ext->mode_context, mbmi->ref_frame, bsize, -1);
      const int16_t newmv_ctx = mode_ctx & NEWMV_CTX_MASK;
      int16_t zeromv_ctx = -1;
      int16_t refmv_ctx = -1;
      if (mbmi->mode != NEWMV) {
        zeromv_ctx = (mode_ctx >> ZEROMV_OFFSET) & ZEROMV_CTX_MASK;
        if (mode_ctx & (1 << ALL_ZERO_FLAG_OFFSET)) {
          assert(mbmi->mode == ZEROMV);
        }
        if (mbmi->mode != ZEROMV) {
          refmv_ctx = (mode_ctx >> REFMV_OFFSET) & REFMV_CTX_MASK;
          if (mode_ctx & (1 << SKIP_NEARESTMV_OFFSET)) refmv_ctx = 6;
          if (mode_ctx & (1 << SKIP_NEARMV_OFFSET)) refmv_ctx = 7;
          if (mode_ctx & (1 << SKIP_NEARESTMV_SUB8X8_OFFSET)) refmv_ctx = 8;
        }
      }

      int8_t ref_frame_type = av1_ref_frame_type(mbmi->ref_frame);
      printf(
          "=== ENCODER ===: "
          "Frame=%d, (mi_row,mi_col)=(%d,%d), mode=%d, bsize=%d, "
          "show_frame=%d, mv[0]=(%d,%d), mv[1]=(%d,%d), ref[0]=%d, "
          "ref[1]=%d, motion_mode=%d, inter_mode_ctx=%d, mode_ctx=%d, "
          "interp_ctx=(%d,%d), interp_filter=(%d,%d), newmv_ctx=%d, "
          "zeromv_ctx=%d, refmv_ctx=%d\n",
          cm->current_video_frame, mi_row, mi_col, mbmi->mode, bsize,
          cm->show_frame, mv[0].as_mv.row, mv[0].as_mv.col, mv[1].as_mv.row,
          mv[1].as_mv.col, mbmi->ref_frame[0], mbmi->ref_frame[1],
          mbmi->motion_mode, mbmi_ext->mode_context[ref_frame_type], mode_ctx,
          interp_ctx[0], interp_ctx[1], interp_filter[0], interp_filter[1],
          newmv_ctx, zeromv_ctx, refmv_ctx);
    }
  }
}
#endif  // ENC_MISMATCH_DEBUG

static void write_mbmi_b(AV1_COMP *cpi, const TileInfo *const tile,
                         aom_writer *w,
#if CONFIG_SUPERTX
                         int supertx_enabled,
#endif
                         int mi_row, int mi_col) {
  AV1_COMMON *const cm = &cpi->common;
  MACROBLOCKD *const xd = &cpi->td.mb.e_mbd;
  MODE_INFO *m;
  int bh, bw;
  xd->mi = cm->mi_grid_visible + (mi_row * cm->mi_stride + mi_col);
  m = xd->mi[0];

  assert(m->mbmi.sb_type <= cm->sb_size ||
         (m->mbmi.sb_type >= BLOCK_4X16 && m->mbmi.sb_type <= BLOCK_32X8));

  bh = mi_size_high[m->mbmi.sb_type];
  bw = mi_size_wide[m->mbmi.sb_type];

  cpi->td.mb.mbmi_ext = cpi->mbmi_ext_base + (mi_row * cm->mi_cols + mi_col);

  set_mi_row_col(xd, tile, mi_row, bh, mi_col, bw,
#if CONFIG_DEPENDENT_HORZTILES
                 cm->dependent_horz_tiles,
#endif  // CONFIG_DEPENDENT_HORZTILES
                 cm->mi_rows, cm->mi_cols);

  if (frame_is_intra_only(cm)) {
    write_mb_modes_kf(cm, xd,
#if CONFIG_INTRABC
                      cpi->td.mb.mbmi_ext,
#endif  // CONFIG_INTRABC
                      mi_row, mi_col, w);
  } else {
#if CONFIG_VAR_TX
    xd->above_txfm_context =
        cm->above_txfm_context + (mi_col << TX_UNIT_WIDE_LOG2);
    xd->left_txfm_context = xd->left_txfm_context_buffer +
                            ((mi_row & MAX_MIB_MASK) << TX_UNIT_HIGH_LOG2);
#endif
#if CONFIG_DUAL_FILTER || CONFIG_WARPED_MOTION
    // has_subpel_mv_component needs the ref frame buffers set up to look
    // up if they are scaled. has_subpel_mv_component is in turn needed by
    // write_switchable_interp_filter, which is called by pack_inter_mode_mvs.
    set_ref_ptrs(cm, xd, m->mbmi.ref_frame[0], m->mbmi.ref_frame[1]);
#if CONFIG_EXT_INTER && CONFIG_COMPOUND_SINGLEREF
    if (!has_second_ref(&m->mbmi) && is_inter_singleref_comp_mode(m->mbmi.mode))
      xd->block_refs[1] = xd->block_refs[0];
#endif  // CONFIG_EXT_INTER && CONFIG_COMPOUND_SINGLEREF
#endif  // CONFIG_DUAL_FILTER || CONFIG_WARPED_MOTION

#if ENC_MISMATCH_DEBUG
    // NOTE(zoeliu): For debug
    enc_dump_logs(cpi, mi_row, mi_col);
#endif  // ENC_MISMATCH_DEBUG

    pack_inter_mode_mvs(cpi, mi_row, mi_col,
#if CONFIG_SUPERTX
                        supertx_enabled,
#endif
                        w);
  }
}

static void write_tokens_b(AV1_COMP *cpi, const TileInfo *const tile,
                           aom_writer *w, const TOKENEXTRA **tok,
                           const TOKENEXTRA *const tok_end, int mi_row,
                           int mi_col) {
  AV1_COMMON *const cm = &cpi->common;
  MACROBLOCKD *const xd = &cpi->td.mb.e_mbd;
  const int mi_offset = mi_row * cm->mi_stride + mi_col;
  MODE_INFO *const m = *(cm->mi_grid_visible + mi_offset);
  MB_MODE_INFO *const mbmi = &m->mbmi;
  int plane;
  int bh, bw;
#if CONFIG_PVQ || CONFIG_LV_MAP
  MACROBLOCK *const x = &cpi->td.mb;
  (void)tok;
  (void)tok_end;
#endif
  xd->mi = cm->mi_grid_visible + mi_offset;

  assert(mbmi->sb_type <= cm->sb_size ||
         (mbmi->sb_type >= BLOCK_4X16 && mbmi->sb_type <= BLOCK_32X8));

  bh = mi_size_high[mbmi->sb_type];
  bw = mi_size_wide[mbmi->sb_type];
  cpi->td.mb.mbmi_ext = cpi->mbmi_ext_base + (mi_row * cm->mi_cols + mi_col);

  set_mi_row_col(xd, tile, mi_row, bh, mi_col, bw,
#if CONFIG_DEPENDENT_HORZTILES
                 cm->dependent_horz_tiles,
#endif  // CONFIG_DEPENDENT_HORZTILES
                 cm->mi_rows, cm->mi_cols);

#if CONFIG_PALETTE
  for (plane = 0; plane <= 1; ++plane) {
    const uint8_t palette_size_plane =
        mbmi->palette_mode_info.palette_size[plane];
    if (palette_size_plane > 0) {
#if CONFIG_INTRABC
      assert(mbmi->use_intrabc == 0);
#endif
      int rows, cols;
      assert(mbmi->sb_type >= BLOCK_8X8);
      av1_get_block_dimensions(mbmi->sb_type, plane, xd, NULL, NULL, &rows,
                               &cols);
      assert(*tok < tok_end);
      pack_palette_tokens(w, tok, palette_size_plane, rows * cols);
      assert(*tok < tok_end + mbmi->skip);
    }
  }
#endif  // CONFIG_PALETTE

#if CONFIG_COEF_INTERLEAVE
  if (!mbmi->skip) {
    const struct macroblockd_plane *const pd_y = &xd->plane[0];
    const struct macroblockd_plane *const pd_c = &xd->plane[1];
    const TX_SIZE tx_log2_y = mbmi->tx_size;
    const TX_SIZE tx_log2_c = av1_get_uv_tx_size(mbmi, pd_c);
    const int tx_sz_y = (1 << tx_log2_y);
    const int tx_sz_c = (1 << tx_log2_c);

    const BLOCK_SIZE plane_bsize_y =
        get_plane_block_size(AOMMAX(mbmi->sb_type, 3), pd_y);
    const BLOCK_SIZE plane_bsize_c =
        get_plane_block_size(AOMMAX(mbmi->sb_type, 3), pd_c);

    const int num_4x4_w_y = num_4x4_blocks_wide_lookup[plane_bsize_y];
    const int num_4x4_w_c = num_4x4_blocks_wide_lookup[plane_bsize_c];
    const int num_4x4_h_y = num_4x4_blocks_high_lookup[plane_bsize_y];
    const int num_4x4_h_c = num_4x4_blocks_high_lookup[plane_bsize_c];

    const int max_4x4_w_y = get_max_4x4_size(num_4x4_w_y, xd->mb_to_right_edge,
                                             pd_y->subsampling_x);
    const int max_4x4_h_y = get_max_4x4_size(num_4x4_h_y, xd->mb_to_bottom_edge,
                                             pd_y->subsampling_y);
    const int max_4x4_w_c = get_max_4x4_size(num_4x4_w_c, xd->mb_to_right_edge,
                                             pd_c->subsampling_x);
    const int max_4x4_h_c = get_max_4x4_size(num_4x4_h_c, xd->mb_to_bottom_edge,
                                             pd_c->subsampling_y);

    // The max_4x4_w/h may be smaller than tx_sz under some corner cases,
    // i.e. when the SB is splitted by tile boundaries.
    const int tu_num_w_y = (max_4x4_w_y + tx_sz_y - 1) / tx_sz_y;
    const int tu_num_h_y = (max_4x4_h_y + tx_sz_y - 1) / tx_sz_y;
    const int tu_num_w_c = (max_4x4_w_c + tx_sz_c - 1) / tx_sz_c;
    const int tu_num_h_c = (max_4x4_h_c + tx_sz_c - 1) / tx_sz_c;
    const int tu_num_y = tu_num_w_y * tu_num_h_y;
    const int tu_num_c = tu_num_w_c * tu_num_h_c;

    int tu_idx_y = 0, tu_idx_c = 0;
    TOKEN_STATS token_stats;
    init_token_stats(&token_stats);

    assert(*tok < tok_end);

    while (tu_idx_y < tu_num_y) {
      pack_mb_tokens(w, tok, tok_end, cm->bit_depth, tx_log2_y, &token_stats);
      assert(*tok < tok_end && (*tok)->token == EOSB_TOKEN);
      (*tok)++;
      tu_idx_y++;

      if (tu_idx_c < tu_num_c) {
        pack_mb_tokens(w, tok, tok_end, cm->bit_depth, tx_log2_c, &token_stats);
        assert(*tok < tok_end && (*tok)->token == EOSB_TOKEN);
        (*tok)++;

        pack_mb_tokens(w, tok, tok_end, cm->bit_depth, tx_log2_c, &token_stats);
        assert(*tok < tok_end && (*tok)->token == EOSB_TOKEN);
        (*tok)++;

        tu_idx_c++;
      }
    }

    // In 422 case, it's possilbe that Chroma has more TUs than Luma
    while (tu_idx_c < tu_num_c) {
      pack_mb_tokens(w, tok, tok_end, cm->bit_depth, tx_log2_c, &token_stats);
      assert(*tok < tok_end && (*tok)->token == EOSB_TOKEN);
      (*tok)++;

      pack_mb_tokens(w, tok, tok_end, cm->bit_depth, tx_log2_c, &token_stats);
      assert(*tok < tok_end && (*tok)->token == EOSB_TOKEN);
      (*tok)++;

      tu_idx_c++;
    }
  }
#else  // CONFIG_COEF_INTERLEAVE
  if (!mbmi->skip) {
#if !CONFIG_PVQ && !CONFIG_LV_MAP
    assert(*tok < tok_end);
#endif
    for (plane = 0; plane < MAX_MB_PLANE; ++plane) {
#if CONFIG_CB4X4
      if (!is_chroma_reference(mi_row, mi_col, mbmi->sb_type,
                               xd->plane[plane].subsampling_x,
                               xd->plane[plane].subsampling_y)) {
        (*tok)++;
        continue;
      }
#endif
#if CONFIG_VAR_TX
      const struct macroblockd_plane *const pd = &xd->plane[plane];
      BLOCK_SIZE bsize = mbmi->sb_type;
#if CONFIG_CHROMA_SUB8X8
      const BLOCK_SIZE plane_bsize =
          AOMMAX(BLOCK_4X4, get_plane_block_size(bsize, pd));
#elif CONFIG_CB4X4
      const BLOCK_SIZE plane_bsize = get_plane_block_size(bsize, pd);
#else
      const BLOCK_SIZE plane_bsize =
          get_plane_block_size(AOMMAX(bsize, BLOCK_8X8), pd);
#endif

      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_wide_log2[0];
      int row, col;
      TOKEN_STATS token_stats;
      init_token_stats(&token_stats);

      const BLOCK_SIZE max_unit_bsize = get_plane_block_size(BLOCK_64X64, pd);
      int mu_blocks_wide =
          block_size_wide[max_unit_bsize] >> tx_size_wide_log2[0];
      int mu_blocks_high =
          block_size_high[max_unit_bsize] >> tx_size_high_log2[0];

      mu_blocks_wide = AOMMIN(num_4x4_w, mu_blocks_wide);
      mu_blocks_high = AOMMIN(num_4x4_h, mu_blocks_high);

      if (is_inter_block(mbmi)) {
        const TX_SIZE max_tx_size = get_vartx_max_txsize(mbmi, plane_bsize);
        int block = 0;
        const int step =
            tx_size_wide_unit[max_tx_size] * tx_size_high_unit[max_tx_size];
        const int bkw = tx_size_wide_unit[max_tx_size];
        const int bkh = tx_size_high_unit[max_tx_size];
        for (row = 0; row < num_4x4_h; row += mu_blocks_high) {
          const int unit_height = AOMMIN(mu_blocks_high + row, num_4x4_h);
          for (col = 0; col < num_4x4_w; col += mu_blocks_wide) {
            int blk_row, blk_col;
            const int unit_width = AOMMIN(mu_blocks_wide + col, num_4x4_w);
            for (blk_row = row; blk_row < unit_height; blk_row += bkh) {
              for (blk_col = col; blk_col < unit_width; blk_col += bkw) {
                pack_txb_tokens(w,
#if CONFIG_LV_MAP
                                cm,
#endif
                                tok, tok_end,
#if CONFIG_PVQ || CONFIG_LV_MAP
                                x,
#endif
                                xd, mbmi, plane, plane_bsize, cm->bit_depth,
                                block, blk_row, blk_col, max_tx_size,
                                &token_stats);
                block += step;
              }
            }
          }
        }
#if CONFIG_RD_DEBUG
        if (mbmi->sb_type >= BLOCK_8X8 &&
            rd_token_stats_mismatch(&mbmi->rd_stats, &token_stats, plane)) {
          dump_mode_info(m);
          assert(0);
        }
#endif  // CONFIG_RD_DEBUG
      } else {
#if CONFIG_LV_MAP
        av1_write_coeffs_mb(cm, x, w, plane);
#else
        const TX_SIZE tx = av1_get_tx_size(plane, xd);
        const int bkw = tx_size_wide_unit[tx];
        const int bkh = tx_size_high_unit[tx];
        int blk_row, blk_col;

        for (row = 0; row < num_4x4_h; row += mu_blocks_high) {
          for (col = 0; col < num_4x4_w; col += mu_blocks_wide) {
            const int unit_height = AOMMIN(mu_blocks_high + row, num_4x4_h);
            const int unit_width = AOMMIN(mu_blocks_wide + col, num_4x4_w);

            for (blk_row = row; blk_row < unit_height; blk_row += bkh) {
              for (blk_col = col; blk_col < unit_width; blk_col += bkw) {
#if !CONFIG_PVQ
                pack_mb_tokens(w, tok, tok_end, cm->bit_depth, tx,
                               &token_stats);
#else
                pack_pvq_tokens(w, x, xd, plane, bsize, tx);
#endif
              }
            }
          }
        }
#endif  // CONFIG_LV_MAP
      }
#else
      const TX_SIZE tx = av1_get_tx_size(plane, xd);
      TOKEN_STATS token_stats;
#if !CONFIG_PVQ
      init_token_stats(&token_stats);
#if CONFIG_LV_MAP
      (void)tx;
      av1_write_coeffs_mb(cm, x, w, plane);
#else   // CONFIG_LV_MAP
      pack_mb_tokens(w, tok, tok_end, cm->bit_depth, tx, &token_stats);
#endif  // CONFIG_LV_MAP

#else
      (void)token_stats;
      pack_pvq_tokens(w, x, xd, plane, mbmi->sb_type, tx);
#endif
#if CONFIG_RD_DEBUG
      if (is_inter_block(mbmi) && mbmi->sb_type >= BLOCK_8X8 &&
          rd_token_stats_mismatch(&mbmi->rd_stats, &token_stats, plane)) {
        dump_mode_info(m);
        assert(0);
      }
#endif  // CONFIG_RD_DEBUG
#endif  // CONFIG_VAR_TX

#if !CONFIG_PVQ && !CONFIG_LV_MAP
      assert(*tok < tok_end && (*tok)->token == EOSB_TOKEN);
      (*tok)++;
#endif
    }
  }
#endif  // CONFIG_COEF_INTERLEAVE
}

#if CONFIG_MOTION_VAR && (CONFIG_NCOBMC || CONFIG_NCOBMC_ADAPT_WEIGHT)
static void write_tokens_sb(AV1_COMP *cpi, const TileInfo *const tile,
                            aom_writer *w, const TOKENEXTRA **tok,
                            const TOKENEXTRA *const tok_end, int mi_row,
                            int mi_col, BLOCK_SIZE bsize) {
  const AV1_COMMON *const cm = &cpi->common;
  const int hbs = mi_size_wide[bsize] / 2;
  PARTITION_TYPE partition;
  BLOCK_SIZE subsize;
#if CONFIG_CB4X4
  const int unify_bsize = 1;
#else
  const int unify_bsize = 0;
#endif

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

  partition = get_partition(cm, mi_row, mi_col, bsize);
  subsize = get_subsize(bsize, partition);

  if (subsize < BLOCK_8X8 && !unify_bsize) {
    write_tokens_b(cpi, tile, w, tok, tok_end, mi_row, mi_col);
  } else {
    switch (partition) {
      case PARTITION_NONE:
        write_tokens_b(cpi, tile, w, tok, tok_end, mi_row, mi_col);
        break;
      case PARTITION_HORZ:
        write_tokens_b(cpi, tile, w, tok, tok_end, mi_row, mi_col);
        if (mi_row + hbs < cm->mi_rows)
          write_tokens_b(cpi, tile, w, tok, tok_end, mi_row + hbs, mi_col);
        break;
      case PARTITION_VERT:
        write_tokens_b(cpi, tile, w, tok, tok_end, mi_row, mi_col);
        if (mi_col + hbs < cm->mi_cols)
          write_tokens_b(cpi, tile, w, tok, tok_end, mi_row, mi_col + hbs);
        break;
      case PARTITION_SPLIT:
        write_tokens_sb(cpi, tile, w, tok, tok_end, mi_row, mi_col, subsize);
        write_tokens_sb(cpi, tile, w, tok, tok_end, mi_row, mi_col + hbs,
                        subsize);
        write_tokens_sb(cpi, tile, w, tok, tok_end, mi_row + hbs, mi_col,
                        subsize);
        write_tokens_sb(cpi, tile, w, tok, tok_end, mi_row + hbs, mi_col + hbs,
                        subsize);
        break;
#if CONFIG_EXT_PARTITION_TYPES
      case PARTITION_HORZ_A:
        write_tokens_b(cpi, tile, w, tok, tok_end, mi_row, mi_col);
        write_tokens_b(cpi, tile, w, tok, tok_end, mi_row, mi_col + hbs);
        write_tokens_b(cpi, tile, w, tok, tok_end, mi_row + hbs, mi_col);
        break;
      case PARTITION_HORZ_B:
        write_tokens_b(cpi, tile, w, tok, tok_end, mi_row, mi_col);
        write_tokens_b(cpi, tile, w, tok, tok_end, mi_row + hbs, mi_col);
        write_tokens_b(cpi, tile, w, tok, tok_end, mi_row + hbs, mi_col + hbs);
        break;
      case PARTITION_VERT_A:
        write_tokens_b(cpi, tile, w, tok, tok_end, mi_row, mi_col);
        write_tokens_b(cpi, tile, w, tok, tok_end, mi_row + hbs, mi_col);
        write_tokens_b(cpi, tile, w, tok, tok_end, mi_row, mi_col + hbs);
        break;
      case PARTITION_VERT_B:
        write_tokens_b(cpi, tile, w, tok, tok_end, mi_row, mi_col);
        write_tokens_b(cpi, tile, w, tok, tok_end, mi_row, mi_col + hbs);
        write_tokens_b(cpi, tile, w, tok, tok_end, mi_row + hbs, mi_col + hbs);
        break;
#endif  // CONFIG_EXT_PARTITION_TYPES
      default: assert(0);
    }
  }
}
#endif

static void write_modes_b(AV1_COMP *cpi, const TileInfo *const tile,
                          aom_writer *w, const TOKENEXTRA **tok,
                          const TOKENEXTRA *const tok_end,
#if CONFIG_SUPERTX
                          int supertx_enabled,
#endif
                          int mi_row, int mi_col) {
  write_mbmi_b(cpi, tile, w,
#if CONFIG_SUPERTX
               supertx_enabled,
#endif
               mi_row, mi_col);
#if CONFIG_MOTION_VAR && (CONFIG_NCOBMC || CONFIG_NCOBMC_ADAPT_WEIGHT)
  (void)tok;
  (void)tok_end;
#else
#if !CONFIG_PVQ && CONFIG_SUPERTX
  if (!supertx_enabled)
#endif
    write_tokens_b(cpi, tile, w, tok, tok_end, mi_row, mi_col);
#endif
}

static void write_partition(const AV1_COMMON *const cm,
                            const MACROBLOCKD *const xd, int hbs, int mi_row,
                            int mi_col, PARTITION_TYPE p, BLOCK_SIZE bsize,
                            aom_writer *w) {
  const int has_rows = (mi_row + hbs) < cm->mi_rows;
  const int has_cols = (mi_col + hbs) < cm->mi_cols;
  const int is_partition_point = bsize >= BLOCK_8X8;
  const int ctx = is_partition_point
                      ? partition_plane_context(xd, mi_row, mi_col,
#if CONFIG_UNPOISON_PARTITION_CTX
                                                has_rows, has_cols,
#endif
                                                bsize)
                      : 0;
#if CONFIG_UNPOISON_PARTITION_CTX
  const aom_prob *const probs =
      ctx < PARTITION_CONTEXTS ? cm->fc->partition_prob[ctx] : NULL;
#else
  const aom_prob *const probs = cm->fc->partition_prob[ctx];
#endif

  FRAME_CONTEXT *ec_ctx = xd->tile_ctx;
  (void)cm;

  if (!is_partition_point) return;

  if (has_rows && has_cols) {
#if CONFIG_EXT_PARTITION_TYPES
    if (bsize <= BLOCK_8X8)
      aom_write_symbol(w, p, ec_ctx->partition_cdf[ctx], PARTITION_TYPES);
    else
      aom_write_symbol(w, p, ec_ctx->partition_cdf[ctx], EXT_PARTITION_TYPES);
#else
    aom_write_symbol(w, p, ec_ctx->partition_cdf[ctx], PARTITION_TYPES);
#endif  // CONFIG_EXT_PARTITION_TYPES
  } else if (!has_rows && has_cols) {
    assert(p == PARTITION_SPLIT || p == PARTITION_HORZ);
    aom_write(w, p == PARTITION_SPLIT, probs[1]);
  } else if (has_rows && !has_cols) {
    assert(p == PARTITION_SPLIT || p == PARTITION_VERT);
    aom_write(w, p == PARTITION_SPLIT, probs[2]);
  } else {
    assert(p == PARTITION_SPLIT);
  }
}

#if CONFIG_SUPERTX
#define write_modes_sb_wrapper(cpi, tile, w, tok, tok_end, supertx_enabled,   \
                               mi_row, mi_col, bsize)                         \
  write_modes_sb(cpi, tile, w, tok, tok_end, supertx_enabled, mi_row, mi_col, \
                 bsize)
#else
#define write_modes_sb_wrapper(cpi, tile, w, tok, tok_end, supertx_enabled, \
                               mi_row, mi_col, bsize)                       \
  write_modes_sb(cpi, tile, w, tok, tok_end, mi_row, mi_col, bsize)
#endif  // CONFIG_SUPERTX

static void write_modes_sb(AV1_COMP *const cpi, const TileInfo *const tile,
                           aom_writer *const w, const TOKENEXTRA **tok,
                           const TOKENEXTRA *const tok_end,
#if CONFIG_SUPERTX
                           int supertx_enabled,
#endif
                           int mi_row, int mi_col, BLOCK_SIZE bsize) {
  const AV1_COMMON *const cm = &cpi->common;
  MACROBLOCKD *const xd = &cpi->td.mb.e_mbd;
  const int hbs = mi_size_wide[bsize] / 2;
#if CONFIG_EXT_PARTITION_TYPES
  const int quarter_step = mi_size_wide[bsize] / 4;
  int i;
#endif
  const PARTITION_TYPE partition = get_partition(cm, mi_row, mi_col, bsize);
  const BLOCK_SIZE subsize = get_subsize(bsize, partition);
#if CONFIG_CB4X4
  const int unify_bsize = 1;
#else
  const int unify_bsize = 0;
#endif

#if CONFIG_SUPERTX
  const int mi_offset = mi_row * cm->mi_stride + mi_col;
  MB_MODE_INFO *mbmi;
  const int pack_token = !supertx_enabled;
  TX_SIZE supertx_size;
  int plane;
#endif

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

  write_partition(cm, xd, hbs, mi_row, mi_col, partition, bsize, w);
#if CONFIG_SUPERTX
  mbmi = &cm->mi_grid_visible[mi_offset]->mbmi;
  xd->mi = cm->mi_grid_visible + mi_offset;
  set_mi_row_col(xd, tile, mi_row, mi_size_high[bsize], mi_col,
                 mi_size_wide[bsize],
#if CONFIG_DEPENDENT_HORZTILES
                 cm->dependent_horz_tiles,
#endif  // CONFIG_DEPENDENT_HORZTILES
                 cm->mi_rows, cm->mi_cols);
  if (!supertx_enabled && !frame_is_intra_only(cm) &&
      partition != PARTITION_NONE && bsize <= MAX_SUPERTX_BLOCK_SIZE &&
      !xd->lossless[0]) {
    aom_prob prob;
    supertx_size = max_txsize_lookup[bsize];
    prob = cm->fc->supertx_prob[partition_supertx_context_lookup[partition]]
                               [supertx_size];
    supertx_enabled = (xd->mi[0]->mbmi.tx_size == supertx_size);
    aom_write(w, supertx_enabled, prob);
  }
#endif  // CONFIG_SUPERTX
  if (subsize < BLOCK_8X8 && !unify_bsize) {
    write_modes_b_wrapper(cpi, tile, w, tok, tok_end, supertx_enabled, mi_row,
                          mi_col);
  } else {
    switch (partition) {
      case PARTITION_NONE:
        write_modes_b_wrapper(cpi, tile, w, tok, tok_end, supertx_enabled,
                              mi_row, mi_col);
        break;
      case PARTITION_HORZ:
        write_modes_b_wrapper(cpi, tile, w, tok, tok_end, supertx_enabled,
                              mi_row, mi_col);
        if (mi_row + hbs < cm->mi_rows)
          write_modes_b_wrapper(cpi, tile, w, tok, tok_end, supertx_enabled,
                                mi_row + hbs, mi_col);
        break;
      case PARTITION_VERT:
        write_modes_b_wrapper(cpi, tile, w, tok, tok_end, supertx_enabled,
                              mi_row, mi_col);
        if (mi_col + hbs < cm->mi_cols)
          write_modes_b_wrapper(cpi, tile, w, tok, tok_end, supertx_enabled,
                                mi_row, mi_col + hbs);
        break;
      case PARTITION_SPLIT:
        write_modes_sb_wrapper(cpi, tile, w, tok, tok_end, supertx_enabled,
                               mi_row, mi_col, subsize);
        write_modes_sb_wrapper(cpi, tile, w, tok, tok_end, supertx_enabled,
                               mi_row, mi_col + hbs, subsize);
        write_modes_sb_wrapper(cpi, tile, w, tok, tok_end, supertx_enabled,
                               mi_row + hbs, mi_col, subsize);
        write_modes_sb_wrapper(cpi, tile, w, tok, tok_end, supertx_enabled,
                               mi_row + hbs, mi_col + hbs, subsize);
        break;
#if CONFIG_EXT_PARTITION_TYPES
      case PARTITION_HORZ_A:
        write_modes_b_wrapper(cpi, tile, w, tok, tok_end, supertx_enabled,
                              mi_row, mi_col);
        write_modes_b_wrapper(cpi, tile, w, tok, tok_end, supertx_enabled,
                              mi_row, mi_col + hbs);
        write_modes_b_wrapper(cpi, tile, w, tok, tok_end, supertx_enabled,
                              mi_row + hbs, mi_col);
        break;
      case PARTITION_HORZ_B:
        write_modes_b_wrapper(cpi, tile, w, tok, tok_end, supertx_enabled,
                              mi_row, mi_col);
        write_modes_b_wrapper(cpi, tile, w, tok, tok_end, supertx_enabled,
                              mi_row + hbs, mi_col);
        write_modes_b_wrapper(cpi, tile, w, tok, tok_end, supertx_enabled,
                              mi_row + hbs, mi_col + hbs);
        break;
      case PARTITION_VERT_A:
        write_modes_b_wrapper(cpi, tile, w, tok, tok_end, supertx_enabled,
                              mi_row, mi_col);
        write_modes_b_wrapper(cpi, tile, w, tok, tok_end, supertx_enabled,
                              mi_row + hbs, mi_col);
        write_modes_b_wrapper(cpi, tile, w, tok, tok_end, supertx_enabled,
                              mi_row, mi_col + hbs);
        break;
      case PARTITION_VERT_B:
        write_modes_b_wrapper(cpi, tile, w, tok, tok_end, supertx_enabled,
                              mi_row, mi_col);
        write_modes_b_wrapper(cpi, tile, w, tok, tok_end, supertx_enabled,
                              mi_row, mi_col + hbs);
        write_modes_b_wrapper(cpi, tile, w, tok, tok_end, supertx_enabled,
                              mi_row + hbs, mi_col + hbs);
        break;
      case PARTITION_HORZ_4:
        for (i = 0; i < 4; ++i) {
          int this_mi_row = mi_row + i * quarter_step;
          if (i > 0 && this_mi_row >= cm->mi_rows) break;

          write_modes_b_wrapper(cpi, tile, w, tok, tok_end, supertx_enabled,
                                this_mi_row, mi_col);
        }
        break;
      case PARTITION_VERT_4:
        for (i = 0; i < 4; ++i) {
          int this_mi_col = mi_col + i * quarter_step;
          if (i > 0 && this_mi_col >= cm->mi_cols) break;

          write_modes_b_wrapper(cpi, tile, w, tok, tok_end, supertx_enabled,
                                mi_row, this_mi_col);
        }
        break;
#endif  // CONFIG_EXT_PARTITION_TYPES
      default: assert(0);
    }
  }
#if CONFIG_SUPERTX
  if (partition != PARTITION_NONE && supertx_enabled && pack_token) {
    int skip;
    const int bsw = mi_size_wide[bsize];
    const int bsh = mi_size_high[bsize];

    xd->mi = cm->mi_grid_visible + mi_offset;
    supertx_size = mbmi->tx_size;
    set_mi_row_col(xd, tile, mi_row, bsh, mi_col, bsw,
#if CONFIG_DEPENDENT_HORZTILES
                   cm->dependent_horz_tiles,
#endif  // CONFIG_DEPENDENT_HORZTILES
                   cm->mi_rows, cm->mi_cols);

    assert(IMPLIES(!cm->seg.enabled, mbmi->segment_id_supertx == 0));
    assert(mbmi->segment_id_supertx < MAX_SEGMENTS);

    skip = write_skip(cm, xd, mbmi->segment_id_supertx, xd->mi[0], w);

    FRAME_CONTEXT *ec_ctx = xd->tile_ctx;

#if CONFIG_EXT_TX
    if (get_ext_tx_types(supertx_size, bsize, 1, cm->reduced_tx_set_used) > 1 &&
        !skip) {
      const int eset =
          get_ext_tx_set(supertx_size, bsize, 1, cm->reduced_tx_set_used);
      if (eset > 0) {
        aom_write_symbol(w, av1_ext_tx_inter_ind[eset][mbmi->tx_type],
                         ec_ctx->inter_ext_tx_cdf[eset][supertx_size],
                         ext_tx_cnt_inter[eset]);
      }
    }
#else
    if (supertx_size < TX_32X32 && !skip) {
      aom_write_symbol(w, mbmi->tx_type, ec_ctx->inter_ext_tx_cdf[supertx_size],
                       TX_TYPES);
    }
#endif  // CONFIG_EXT_TX

    if (!skip) {
      assert(*tok < tok_end);
      for (plane = 0; plane < MAX_MB_PLANE; ++plane) {
        const struct macroblockd_plane *const pd = &xd->plane[plane];
        const int mbmi_txb_size = txsize_to_bsize[mbmi->tx_size];
        const BLOCK_SIZE plane_bsize = get_plane_block_size(mbmi_txb_size, pd);

        const int max_blocks_wide = max_block_wide(xd, plane_bsize, plane);
        const int max_blocks_high = max_block_high(xd, plane_bsize, plane);

        int row, col;
        const TX_SIZE tx = av1_get_tx_size(plane, xd);
        BLOCK_SIZE txb_size = txsize_to_bsize[tx];

        const int stepr = tx_size_high_unit[txb_size];
        const int stepc = tx_size_wide_unit[txb_size];

        TOKEN_STATS token_stats;
        token_stats.cost = 0;
        for (row = 0; row < max_blocks_high; row += stepr)
          for (col = 0; col < max_blocks_wide; col += stepc)
            pack_mb_tokens(w, tok, tok_end, cm->bit_depth, tx, &token_stats);
        assert(*tok < tok_end && (*tok)->token == EOSB_TOKEN);
        (*tok)++;
      }
    }
#if CONFIG_VAR_TX
    xd->above_txfm_context = cm->above_txfm_context + mi_col;
    xd->left_txfm_context =
        xd->left_txfm_context_buffer + (mi_row & MAX_MIB_MASK);
    set_txfm_ctxs(xd->mi[0]->mbmi.tx_size, bsw, bsh, skip, xd);
#endif
  }
#endif  // CONFIG_SUPERTX

// update partition context
#if CONFIG_EXT_PARTITION_TYPES
  update_ext_partition_context(xd, mi_row, mi_col, subsize, bsize, partition);
#else
  if (bsize >= BLOCK_8X8 &&
      (bsize == BLOCK_8X8 || partition != PARTITION_SPLIT))
    update_partition_context(xd, mi_row, mi_col, subsize, bsize);
#endif  // CONFIG_EXT_PARTITION_TYPES

#if CONFIG_CDEF
  if (bsize == cm->sb_size && cm->cdef_bits != 0 && !cm->all_lossless) {
    int width_step = mi_size_wide[BLOCK_64X64];
    int height_step = mi_size_high[BLOCK_64X64];
    int width, height;
    for (height = 0; (height < mi_size_high[cm->sb_size]) &&
                     (mi_row + height < cm->mi_rows);
         height += height_step) {
      for (width = 0; (width < mi_size_wide[cm->sb_size]) &&
                      (mi_col + width < cm->mi_cols);
           width += width_step) {
        if (!sb_all_skip(cm, mi_row + height, mi_col + width))
          aom_write_literal(
              w, cm->mi_grid_visible[(mi_row + height) * cm->mi_stride +
                                     (mi_col + width)]
                     ->mbmi.cdef_strength,
              cm->cdef_bits);
      }
    }
  }
#endif
}

static void write_modes(AV1_COMP *const cpi, const TileInfo *const tile,
                        aom_writer *const w, const TOKENEXTRA **tok,
                        const TOKENEXTRA *const tok_end) {
  AV1_COMMON *const cm = &cpi->common;
  MACROBLOCKD *const xd = &cpi->td.mb.e_mbd;
  const int mi_row_start = tile->mi_row_start;
  const int mi_row_end = tile->mi_row_end;
  const int mi_col_start = tile->mi_col_start;
  const int mi_col_end = tile->mi_col_end;
  int mi_row, mi_col;

#if CONFIG_DEPENDENT_HORZTILES
  if (!cm->dependent_horz_tiles || mi_row_start == 0 ||
      tile->tg_horz_boundary) {
    av1_zero_above_context(cm, mi_col_start, mi_col_end);
  }
#else
  av1_zero_above_context(cm, mi_col_start, mi_col_end);
#endif
#if CONFIG_PVQ
  assert(cpi->td.mb.pvq_q->curr_pos == 0);
#endif
#if CONFIG_DELTA_Q
  if (cpi->common.delta_q_present_flag) {
    xd->prev_qindex = cpi->common.base_qindex;
#if CONFIG_EXT_DELTA_Q
    if (cpi->common.delta_lf_present_flag) {
      xd->prev_delta_lf_from_base = 0;
    }
#endif  // CONFIG_EXT_DELTA_Q
  }
#endif

  for (mi_row = mi_row_start; mi_row < mi_row_end; mi_row += cm->mib_size) {
    av1_zero_left_context(xd);

    for (mi_col = mi_col_start; mi_col < mi_col_end; mi_col += cm->mib_size) {
      write_modes_sb_wrapper(cpi, tile, w, tok, tok_end, 0, mi_row, mi_col,
                             cm->sb_size);
#if CONFIG_MOTION_VAR && (CONFIG_NCOBMC || CONFIG_NCOBMC_ADAPT_WEIGHT)
      write_tokens_sb(cpi, tile, w, tok, tok_end, mi_row, mi_col, cm->sb_size);
#endif
    }
  }
#if CONFIG_PVQ
  // Check that the number of PVQ blocks encoded and written to the bitstream
  // are the same
  assert(cpi->td.mb.pvq_q->curr_pos == cpi->td.mb.pvq_q->last_pos);
  // Reset curr_pos in case we repack the bitstream
  cpi->td.mb.pvq_q->curr_pos = 0;
#endif
}

#if CONFIG_LOOP_RESTORATION
static void encode_restoration_mode(AV1_COMMON *cm,
                                    struct aom_write_bit_buffer *wb) {
  int p;
  RestorationInfo *rsi = &cm->rst_info[0];
  switch (rsi->frame_restoration_type) {
    case RESTORE_NONE:
      aom_wb_write_bit(wb, 0);
      aom_wb_write_bit(wb, 0);
      break;
    case RESTORE_WIENER:
      aom_wb_write_bit(wb, 1);
      aom_wb_write_bit(wb, 0);
      break;
    case RESTORE_SGRPROJ:
      aom_wb_write_bit(wb, 1);
      aom_wb_write_bit(wb, 1);
      break;
    case RESTORE_SWITCHABLE:
      aom_wb_write_bit(wb, 0);
      aom_wb_write_bit(wb, 1);
      break;
    default: assert(0);
  }
  for (p = 1; p < MAX_MB_PLANE; ++p) {
    rsi = &cm->rst_info[p];
    switch (rsi->frame_restoration_type) {
      case RESTORE_NONE: aom_wb_write_bit(wb, 0); break;
      case RESTORE_WIENER:
        aom_wb_write_bit(wb, 1);
        aom_wb_write_bit(wb, 0);
        break;
      case RESTORE_SGRPROJ:
        aom_wb_write_bit(wb, 1);
        aom_wb_write_bit(wb, 1);
        break;
      default: assert(0);
    }
  }
  if (cm->rst_info[0].frame_restoration_type != RESTORE_NONE ||
      cm->rst_info[1].frame_restoration_type != RESTORE_NONE ||
      cm->rst_info[2].frame_restoration_type != RESTORE_NONE) {
    rsi = &cm->rst_info[0];
    aom_wb_write_bit(wb, rsi->restoration_tilesize != RESTORATION_TILESIZE_MAX);
    if (rsi->restoration_tilesize != RESTORATION_TILESIZE_MAX) {
      aom_wb_write_bit(
          wb, rsi->restoration_tilesize != (RESTORATION_TILESIZE_MAX >> 1));
    }
  }
  int s = AOMMIN(cm->subsampling_x, cm->subsampling_y);
  if (s && (cm->rst_info[1].frame_restoration_type != RESTORE_NONE ||
            cm->rst_info[2].frame_restoration_type != RESTORE_NONE)) {
    aom_wb_write_bit(wb, cm->rst_info[1].restoration_tilesize !=
                             cm->rst_info[0].restoration_tilesize);
    assert(cm->rst_info[1].restoration_tilesize ==
               cm->rst_info[0].restoration_tilesize ||
           cm->rst_info[1].restoration_tilesize ==
               (cm->rst_info[0].restoration_tilesize >> s));
    assert(cm->rst_info[2].restoration_tilesize ==
           cm->rst_info[1].restoration_tilesize);
  } else if (!s) {
    assert(cm->rst_info[1].restoration_tilesize ==
           cm->rst_info[0].restoration_tilesize);
    assert(cm->rst_info[2].restoration_tilesize ==
           cm->rst_info[1].restoration_tilesize);
  }
}

static void write_wiener_filter(WienerInfo *wiener_info,
                                WienerInfo *ref_wiener_info, aom_writer *wb) {
  aom_write_primitive_refsubexpfin(
      wb, WIENER_FILT_TAP0_MAXV - WIENER_FILT_TAP0_MINV + 1,
      WIENER_FILT_TAP0_SUBEXP_K,
      ref_wiener_info->vfilter[0] - WIENER_FILT_TAP0_MINV,
      wiener_info->vfilter[0] - WIENER_FILT_TAP0_MINV);
  aom_write_primitive_refsubexpfin(
      wb, WIENER_FILT_TAP1_MAXV - WIENER_FILT_TAP1_MINV + 1,
      WIENER_FILT_TAP1_SUBEXP_K,
      ref_wiener_info->vfilter[1] - WIENER_FILT_TAP1_MINV,
      wiener_info->vfilter[1] - WIENER_FILT_TAP1_MINV);
  aom_write_primitive_refsubexpfin(
      wb, WIENER_FILT_TAP2_MAXV - WIENER_FILT_TAP2_MINV + 1,
      WIENER_FILT_TAP2_SUBEXP_K,
      ref_wiener_info->vfilter[2] - WIENER_FILT_TAP2_MINV,
      wiener_info->vfilter[2] - WIENER_FILT_TAP2_MINV);
  aom_write_primitive_refsubexpfin(
      wb, WIENER_FILT_TAP0_MAXV - WIENER_FILT_TAP0_MINV + 1,
      WIENER_FILT_TAP0_SUBEXP_K,
      ref_wiener_info->hfilter[0] - WIENER_FILT_TAP0_MINV,
      wiener_info->hfilter[0] - WIENER_FILT_TAP0_MINV);
  aom_write_primitive_refsubexpfin(
      wb, WIENER_FILT_TAP1_MAXV - WIENER_FILT_TAP1_MINV + 1,
      WIENER_FILT_TAP1_SUBEXP_K,
      ref_wiener_info->hfilter[1] - WIENER_FILT_TAP1_MINV,
      wiener_info->hfilter[1] - WIENER_FILT_TAP1_MINV);
  aom_write_primitive_refsubexpfin(
      wb, WIENER_FILT_TAP2_MAXV - WIENER_FILT_TAP2_MINV + 1,
      WIENER_FILT_TAP2_SUBEXP_K,
      ref_wiener_info->hfilter[2] - WIENER_FILT_TAP2_MINV,
      wiener_info->hfilter[2] - WIENER_FILT_TAP2_MINV);
  memcpy(ref_wiener_info, wiener_info, sizeof(*wiener_info));
}

static void write_sgrproj_filter(SgrprojInfo *sgrproj_info,
                                 SgrprojInfo *ref_sgrproj_info,
                                 aom_writer *wb) {
  aom_write_literal(wb, sgrproj_info->ep, SGRPROJ_PARAMS_BITS);
  aom_write_primitive_refsubexpfin(wb, SGRPROJ_PRJ_MAX0 - SGRPROJ_PRJ_MIN0 + 1,
                                   SGRPROJ_PRJ_SUBEXP_K,
                                   ref_sgrproj_info->xqd[0] - SGRPROJ_PRJ_MIN0,
                                   sgrproj_info->xqd[0] - SGRPROJ_PRJ_MIN0);
  aom_write_primitive_refsubexpfin(wb, SGRPROJ_PRJ_MAX1 - SGRPROJ_PRJ_MIN1 + 1,
                                   SGRPROJ_PRJ_SUBEXP_K,
                                   ref_sgrproj_info->xqd[1] - SGRPROJ_PRJ_MIN1,
                                   sgrproj_info->xqd[1] - SGRPROJ_PRJ_MIN1);
  memcpy(ref_sgrproj_info, sgrproj_info, sizeof(*sgrproj_info));
}

static void encode_restoration(AV1_COMMON *cm, aom_writer *wb) {
  int i, p;
#if CONFIG_FRAME_SUPERRES
  const int width = cm->superres_upscaled_width;
  const int height = cm->superres_upscaled_height;
#else
  const int width = cm->width;
  const int height = cm->height;
#endif  // CONFIG_FRAME_SUPERRES
  const int ntiles =
      av1_get_rest_ntiles(width, height, cm->rst_info[0].restoration_tilesize,
                          NULL, NULL, NULL, NULL);
  WienerInfo ref_wiener_info;
  SgrprojInfo ref_sgrproj_info;
  set_default_wiener(&ref_wiener_info);
  set_default_sgrproj(&ref_sgrproj_info);
  const int ntiles_uv = av1_get_rest_ntiles(
      ROUND_POWER_OF_TWO(width, cm->subsampling_x),
      ROUND_POWER_OF_TWO(height, cm->subsampling_y),
      cm->rst_info[1].restoration_tilesize, NULL, NULL, NULL, NULL);
  RestorationInfo *rsi = &cm->rst_info[0];
  if (rsi->frame_restoration_type != RESTORE_NONE) {
    if (rsi->frame_restoration_type == RESTORE_SWITCHABLE) {
      // RESTORE_SWITCHABLE
      for (i = 0; i < ntiles; ++i) {
        av1_write_token(
            wb, av1_switchable_restore_tree, cm->fc->switchable_restore_prob,
            &switchable_restore_encodings[rsi->restoration_type[i]]);
        if (rsi->restoration_type[i] == RESTORE_WIENER) {
          write_wiener_filter(&rsi->wiener_info[i], &ref_wiener_info, wb);
        } else if (rsi->restoration_type[i] == RESTORE_SGRPROJ) {
          write_sgrproj_filter(&rsi->sgrproj_info[i], &ref_sgrproj_info, wb);
        }
      }
    } else if (rsi->frame_restoration_type == RESTORE_WIENER) {
      for (i = 0; i < ntiles; ++i) {
        aom_write(wb, rsi->restoration_type[i] != RESTORE_NONE,
                  RESTORE_NONE_WIENER_PROB);
        if (rsi->restoration_type[i] != RESTORE_NONE) {
          write_wiener_filter(&rsi->wiener_info[i], &ref_wiener_info, wb);
        }
      }
    } else if (rsi->frame_restoration_type == RESTORE_SGRPROJ) {
      for (i = 0; i < ntiles; ++i) {
        aom_write(wb, rsi->restoration_type[i] != RESTORE_NONE,
                  RESTORE_NONE_SGRPROJ_PROB);
        if (rsi->restoration_type[i] != RESTORE_NONE) {
          write_sgrproj_filter(&rsi->sgrproj_info[i], &ref_sgrproj_info, wb);
        }
      }
    }
  }
  for (p = 1; p < MAX_MB_PLANE; ++p) {
    set_default_wiener(&ref_wiener_info);
    set_default_sgrproj(&ref_sgrproj_info);
    rsi = &cm->rst_info[p];
    if (rsi->frame_restoration_type == RESTORE_WIENER) {
      for (i = 0; i < ntiles_uv; ++i) {
        if (ntiles_uv > 1)
          aom_write(wb, rsi->restoration_type[i] != RESTORE_NONE,
                    RESTORE_NONE_WIENER_PROB);
        if (rsi->restoration_type[i] != RESTORE_NONE) {
          write_wiener_filter(&rsi->wiener_info[i], &ref_wiener_info, wb);
        }
      }
    } else if (rsi->frame_restoration_type == RESTORE_SGRPROJ) {
      for (i = 0; i < ntiles_uv; ++i) {
        if (ntiles_uv > 1)
          aom_write(wb, rsi->restoration_type[i] != RESTORE_NONE,
                    RESTORE_NONE_SGRPROJ_PROB);
        if (rsi->restoration_type[i] != RESTORE_NONE) {
          write_sgrproj_filter(&rsi->sgrproj_info[i], &ref_sgrproj_info, wb);
        }
      }
    } else if (rsi->frame_restoration_type != RESTORE_NONE) {
      assert(0);
    }
  }
}
#endif  // CONFIG_LOOP_RESTORATION

static void encode_loopfilter(AV1_COMMON *cm, struct aom_write_bit_buffer *wb) {
  int i;
  struct loopfilter *lf = &cm->lf;

  // Encode the loop filter level and type
  aom_wb_write_literal(wb, lf->filter_level, 6);
#if CONFIG_UV_LVL
  if (lf->filter_level > 0) {
    aom_wb_write_literal(wb, lf->filter_level_u, 6);
    aom_wb_write_literal(wb, lf->filter_level_v, 6);
  }
#endif
  aom_wb_write_literal(wb, lf->sharpness_level, 3);

  // Write out loop filter deltas applied at the MB level based on mode or
  // ref frame (if they are enabled).
  aom_wb_write_bit(wb, lf->mode_ref_delta_enabled);

  if (lf->mode_ref_delta_enabled) {
    aom_wb_write_bit(wb, lf->mode_ref_delta_update);
    if (lf->mode_ref_delta_update) {
      for (i = 0; i < TOTAL_REFS_PER_FRAME; i++) {
        const int delta = lf->ref_deltas[i];
        const int changed = delta != lf->last_ref_deltas[i];
        aom_wb_write_bit(wb, changed);
        if (changed) {
          lf->last_ref_deltas[i] = delta;
          aom_wb_write_inv_signed_literal(wb, delta, 6);
        }
      }

      for (i = 0; i < MAX_MODE_LF_DELTAS; i++) {
        const int delta = lf->mode_deltas[i];
        const int changed = delta != lf->last_mode_deltas[i];
        aom_wb_write_bit(wb, changed);
        if (changed) {
          lf->last_mode_deltas[i] = delta;
          aom_wb_write_inv_signed_literal(wb, delta, 6);
        }
      }
    }
  }
}

#if CONFIG_CDEF
static void encode_cdef(const AV1_COMMON *cm, struct aom_write_bit_buffer *wb) {
  int i;
  aom_wb_write_literal(wb, cm->cdef_dering_damping - 5, 1);
  aom_wb_write_literal(wb, cm->cdef_clpf_damping - 3, 2);
  aom_wb_write_literal(wb, cm->cdef_bits, 2);
  for (i = 0; i < cm->nb_cdef_strengths; i++) {
    aom_wb_write_literal(wb, cm->cdef_strengths[i], CDEF_STRENGTH_BITS);
    aom_wb_write_literal(wb, cm->cdef_uv_strengths[i], CDEF_STRENGTH_BITS);
  }
}
#endif

static void write_delta_q(struct aom_write_bit_buffer *wb, int delta_q) {
  if (delta_q != 0) {
    aom_wb_write_bit(wb, 1);
    aom_wb_write_inv_signed_literal(wb, delta_q, 6);
  } else {
    aom_wb_write_bit(wb, 0);
  }
}

static void encode_quantization(const AV1_COMMON *const cm,
                                struct aom_write_bit_buffer *wb) {
  aom_wb_write_literal(wb, cm->base_qindex, QINDEX_BITS);
  write_delta_q(wb, cm->y_dc_delta_q);
  write_delta_q(wb, cm->uv_dc_delta_q);
  write_delta_q(wb, cm->uv_ac_delta_q);
#if CONFIG_AOM_QM
  aom_wb_write_bit(wb, cm->using_qmatrix);
  if (cm->using_qmatrix) {
    aom_wb_write_literal(wb, cm->min_qmlevel, QM_LEVEL_BITS);
    aom_wb_write_literal(wb, cm->max_qmlevel, QM_LEVEL_BITS);
  }
#endif
}

static void encode_segmentation(AV1_COMMON *cm, MACROBLOCKD *xd,
                                struct aom_write_bit_buffer *wb) {
  int i, j;
  const struct segmentation *seg = &cm->seg;

  aom_wb_write_bit(wb, seg->enabled);
  if (!seg->enabled) return;

  // Segmentation map
  if (!frame_is_intra_only(cm) && !cm->error_resilient_mode) {
    aom_wb_write_bit(wb, seg->update_map);
  } else {
    assert(seg->update_map == 1);
  }
  if (seg->update_map) {
    // Select the coding strategy (temporal or spatial)
    av1_choose_segmap_coding_method(cm, xd);

    // Write out the chosen coding method.
    if (!frame_is_intra_only(cm) && !cm->error_resilient_mode) {
      aom_wb_write_bit(wb, seg->temporal_update);
    } else {
      assert(seg->temporal_update == 0);
    }
  }

  // Segmentation data
  aom_wb_write_bit(wb, seg->update_data);
  if (seg->update_data) {
    aom_wb_write_bit(wb, seg->abs_delta);

    for (i = 0; i < MAX_SEGMENTS; i++) {
      for (j = 0; j < SEG_LVL_MAX; j++) {
        const int active = segfeature_active(seg, i, j);
        aom_wb_write_bit(wb, active);
        if (active) {
          const int data = get_segdata(seg, i, j);
          const int data_max = av1_seg_feature_data_max(j);

          if (av1_is_segfeature_signed(j)) {
            encode_unsigned_max(wb, abs(data), data_max);
            aom_wb_write_bit(wb, data < 0);
          } else {
            encode_unsigned_max(wb, data, data_max);
          }
        }
      }
    }
  }
}

static void write_tx_mode(AV1_COMMON *cm, TX_MODE *mode,
                          struct aom_write_bit_buffer *wb) {
  if (cm->all_lossless) {
    *mode = ONLY_4X4;
    return;
  }
#if CONFIG_VAR_TX_NO_TX_MODE
  (void)wb;
  *mode = TX_MODE_SELECT;
  return;
#else
#if CONFIG_TX64X64
  aom_wb_write_bit(wb, *mode == TX_MODE_SELECT);
  if (*mode != TX_MODE_SELECT) {
    aom_wb_write_literal(wb, AOMMIN(*mode, ALLOW_32X32), 2);
    if (*mode >= ALLOW_32X32) aom_wb_write_bit(wb, *mode == ALLOW_64X64);
  }
#else
  aom_wb_write_bit(wb, *mode == TX_MODE_SELECT);
  if (*mode != TX_MODE_SELECT) aom_wb_write_literal(wb, *mode, 2);
#endif  // CONFIG_TX64X64
#endif  // CONFIG_VAR_TX_NO_TX_MODE
}

static void write_frame_interp_filter(InterpFilter filter,
                                      struct aom_write_bit_buffer *wb) {
  aom_wb_write_bit(wb, filter == SWITCHABLE);
  if (filter != SWITCHABLE)
    aom_wb_write_literal(wb, filter, LOG_SWITCHABLE_FILTERS);
}

static void fix_interp_filter(AV1_COMMON *cm, FRAME_COUNTS *counts) {
  if (cm->interp_filter == SWITCHABLE) {
    // Check to see if only one of the filters is actually used
    int count[SWITCHABLE_FILTERS];
    int i, j, c = 0;
    for (i = 0; i < SWITCHABLE_FILTERS; ++i) {
      count[i] = 0;
      for (j = 0; j < SWITCHABLE_FILTER_CONTEXTS; ++j)
        count[i] += counts->switchable_interp[j][i];
      c += (count[i] > 0);
    }
    if (c == 1) {
      // Only one filter is used. So set the filter at frame level
      for (i = 0; i < SWITCHABLE_FILTERS; ++i) {
        if (count[i]) {
#if CONFIG_MOTION_VAR && (CONFIG_WARPED_MOTION || CONFIG_GLOBAL_MOTION)
#if CONFIG_WARPED_MOTION
          if (i == EIGHTTAP_REGULAR || WARP_WM_NEIGHBORS_WITH_OBMC)
#else
          if (i == EIGHTTAP_REGULAR || WARP_GM_NEIGHBORS_WITH_OBMC)
#endif  // CONFIG_WARPED_MOTION
#endif  // CONFIG_MOTION_VAR && (CONFIG_WARPED_MOTION || CONFIG_GLOBAL_MOTION)
            cm->interp_filter = i;
          break;
        }
      }
    }
  }
}

static void write_tile_info(const AV1_COMMON *const cm,
                            struct aom_write_bit_buffer *wb) {
#if CONFIG_EXT_TILE
  if (cm->large_scale_tile) {
    const int tile_width =
        ALIGN_POWER_OF_TWO(cm->tile_width, cm->mib_size_log2) >>
        cm->mib_size_log2;
    const int tile_height =
        ALIGN_POWER_OF_TWO(cm->tile_height, cm->mib_size_log2) >>
        cm->mib_size_log2;

    assert(tile_width > 0);
    assert(tile_height > 0);

// Write the tile sizes
#if CONFIG_EXT_PARTITION
    if (cm->sb_size == BLOCK_128X128) {
      assert(tile_width <= 32);
      assert(tile_height <= 32);
      aom_wb_write_literal(wb, tile_width - 1, 5);
      aom_wb_write_literal(wb, tile_height - 1, 5);
    } else {
#endif  // CONFIG_EXT_PARTITION
      assert(tile_width <= 64);
      assert(tile_height <= 64);
      aom_wb_write_literal(wb, tile_width - 1, 6);
      aom_wb_write_literal(wb, tile_height - 1, 6);
#if CONFIG_EXT_PARTITION
    }
#endif  // CONFIG_EXT_PARTITION
  } else {
#endif  // CONFIG_EXT_TILE
    int min_log2_tile_cols, max_log2_tile_cols, ones;
    av1_get_tile_n_bits(cm->mi_cols, &min_log2_tile_cols, &max_log2_tile_cols);

    // columns
    ones = cm->log2_tile_cols - min_log2_tile_cols;
    while (ones--) aom_wb_write_bit(wb, 1);

    if (cm->log2_tile_cols < max_log2_tile_cols) aom_wb_write_bit(wb, 0);

    // rows
    aom_wb_write_bit(wb, cm->log2_tile_rows != 0);
    if (cm->log2_tile_rows != 0) aom_wb_write_bit(wb, cm->log2_tile_rows != 1);
#if CONFIG_DEPENDENT_HORZTILES
    if (cm->log2_tile_rows != 0) aom_wb_write_bit(wb, cm->dependent_horz_tiles);
#endif
#if CONFIG_EXT_TILE
  }
#endif  // CONFIG_EXT_TILE

#if CONFIG_LOOPFILTERING_ACROSS_TILES
  aom_wb_write_bit(wb, cm->loop_filter_across_tiles_enabled);
#endif  // CONFIG_LOOPFILTERING_ACROSS_TILES
}

static int get_refresh_mask(AV1_COMP *cpi) {
  int refresh_mask = 0;

#if CONFIG_EXT_REFS
  // NOTE(zoeliu): When LAST_FRAME is to get refreshed, the decoder will be
  // notified to get LAST3_FRAME refreshed and then the virtual indexes for all
  // the 3 LAST reference frames will be updated accordingly, i.e.:
  // (1) The original virtual index for LAST3_FRAME will become the new virtual
  //     index for LAST_FRAME; and
  // (2) The original virtual indexes for LAST_FRAME and LAST2_FRAME will be
  //     shifted and become the new virtual indexes for LAST2_FRAME and
  //     LAST3_FRAME.
  refresh_mask |=
      (cpi->refresh_last_frame << cpi->lst_fb_idxes[LAST_REF_FRAMES - 1]);
  if (cpi->rc.is_bwd_ref_frame && cpi->num_extra_arfs) {
    // We have swapped the virtual indices
    refresh_mask |= (cpi->refresh_bwd_ref_frame << cpi->arf_map[0]);
  } else {
    refresh_mask |= (cpi->refresh_bwd_ref_frame << cpi->bwd_fb_idx);
  }
#else
  refresh_mask |= (cpi->refresh_last_frame << cpi->lst_fb_idx);
#endif  // CONFIG_EXT_REFS

  if (av1_preserve_existing_gf(cpi)) {
    // We have decided to preserve the previously existing golden frame as our
    // new ARF frame. However, in the short term we leave it in the GF slot and,
    // if we're updating the GF with the current decoded frame, we save it
    // instead to the ARF slot.
    // Later, in the function av1_encoder.c:av1_update_reference_frames() we
    // will swap gld_fb_idx and alt_fb_idx to achieve our objective. We do it
    // there so that it can be done outside of the recode loop.
    // Note: This is highly specific to the use of ARF as a forward reference,
    // and this needs to be generalized as other uses are implemented
    // (like RTC/temporal scalability).
    return refresh_mask | (cpi->refresh_golden_frame << cpi->alt_fb_idx);
  } else {
#if CONFIG_EXT_REFS
    const GF_GROUP *const gf_group = &cpi->twopass.gf_group;
    int arf_idx = cpi->arf_map[gf_group->arf_update_idx[gf_group->index]];
#else
    int arf_idx = cpi->alt_fb_idx;
    if ((cpi->oxcf.pass == 2) && cpi->multi_arf_allowed) {
      const GF_GROUP *const gf_group = &cpi->twopass.gf_group;
      arf_idx = gf_group->arf_update_idx[gf_group->index];
    }
#endif  // CONFIG_EXT_REFS
    return refresh_mask | (cpi->refresh_golden_frame << cpi->gld_fb_idx) |
           (cpi->refresh_alt_ref_frame << arf_idx);
  }
}

#if CONFIG_EXT_TILE
static INLINE int find_identical_tile(
    const int tile_row, const int tile_col,
    TileBufferEnc (*const tile_buffers)[1024]) {
  const MV32 candidate_offset[1] = { { 1, 0 } };
  const uint8_t *const cur_tile_data =
      tile_buffers[tile_row][tile_col].data + 4;
  const size_t cur_tile_size = tile_buffers[tile_row][tile_col].size;

  int i;

  if (tile_row == 0) return 0;

  // (TODO: yunqingwang) For now, only above tile is checked and used.
  // More candidates such as left tile can be added later.
  for (i = 0; i < 1; i++) {
    int row_offset = candidate_offset[0].row;
    int col_offset = candidate_offset[0].col;
    int row = tile_row - row_offset;
    int col = tile_col - col_offset;
    uint8_t tile_hdr;
    const uint8_t *tile_data;
    TileBufferEnc *candidate;

    if (row < 0 || col < 0) continue;

    tile_hdr = *(tile_buffers[row][col].data);

    // Read out tcm bit
    if ((tile_hdr >> 7) == 1) {
      // The candidate is a copy tile itself
      row_offset += tile_hdr & 0x7f;
      row = tile_row - row_offset;
    }

    candidate = &tile_buffers[row][col];

    if (row_offset >= 128 || candidate->size != cur_tile_size) continue;

    tile_data = candidate->data + 4;

    if (memcmp(tile_data, cur_tile_data, cur_tile_size) != 0) continue;

    // Identical tile found
    assert(row_offset > 0);
    return row_offset;
  }

  // No identical tile found
  return 0;
}
#endif  // CONFIG_EXT_TILE

static uint32_t write_tiles(AV1_COMP *const cpi, uint8_t *const dst,
                            unsigned int *max_tile_size,
                            unsigned int *max_tile_col_size) {
  const AV1_COMMON *const cm = &cpi->common;
#if CONFIG_ANS
  struct BufAnsCoder *buf_ans = &cpi->buf_ans;
#else
  aom_writer mode_bc;
#endif  // CONFIG_ANS
  int tile_row, tile_col;
  TOKENEXTRA *(*const tok_buffers)[MAX_TILE_COLS] = cpi->tile_tok;
  TileBufferEnc(*const tile_buffers)[MAX_TILE_COLS] = cpi->tile_buffers;
  uint32_t total_size = 0;
  const int tile_cols = cm->tile_cols;
  const int tile_rows = cm->tile_rows;
  unsigned int tile_size = 0;
  const int have_tiles = tile_cols * tile_rows > 1;
  struct aom_write_bit_buffer wb = { dst, 0 };
  const int n_log2_tiles = cm->log2_tile_rows + cm->log2_tile_cols;
  uint32_t comp_hdr_size;
  // Fixed size tile groups for the moment
  const int num_tg_hdrs = cm->num_tg;
  const int tg_size =
#if CONFIG_EXT_TILE
      (cm->large_scale_tile)
          ? 1
          :
#endif  // CONFIG_EXT_TILE
          (tile_rows * tile_cols + num_tg_hdrs - 1) / num_tg_hdrs;
  int tile_count = 0;
  int tg_count = 1;
  int tile_size_bytes = 4;
  int tile_col_size_bytes;
  uint32_t uncompressed_hdr_size = 0;
  struct aom_write_bit_buffer comp_hdr_len_wb;
  struct aom_write_bit_buffer tg_params_wb;
  struct aom_write_bit_buffer tile_size_bytes_wb;
  uint32_t saved_offset;
  int mtu_size = cpi->oxcf.mtu;
  int curr_tg_data_size = 0;
  int hdr_size;

  *max_tile_size = 0;
  *max_tile_col_size = 0;

// All tile size fields are output on 4 bytes. A call to remux_tiles will
// later compact the data if smaller headers are adequate.

#if CONFIG_EXT_TILE
  if (cm->large_scale_tile) {
    for (tile_col = 0; tile_col < tile_cols; tile_col++) {
      TileInfo tile_info;
      const int is_last_col = (tile_col == tile_cols - 1);
      const uint32_t col_offset = total_size;

      av1_tile_set_col(&tile_info, cm, tile_col);

      // The last column does not have a column header
      if (!is_last_col) total_size += 4;

      for (tile_row = 0; tile_row < tile_rows; tile_row++) {
        TileBufferEnc *const buf = &tile_buffers[tile_row][tile_col];
        const TOKENEXTRA *tok = tok_buffers[tile_row][tile_col];
        const TOKENEXTRA *tok_end = tok + cpi->tok_count[tile_row][tile_col];
        const int data_offset = have_tiles ? 4 : 0;
        const int tile_idx = tile_row * tile_cols + tile_col;
        TileDataEnc *this_tile = &cpi->tile_data[tile_idx];
        av1_tile_set_row(&tile_info, cm, tile_row);

        buf->data = dst + total_size;

        // Is CONFIG_EXT_TILE = 1, every tile in the row has a header,
        // even for the last one, unless no tiling is used at all.
        total_size += data_offset;
        // Initialise tile context from the frame context
        this_tile->tctx = *cm->fc;
        cpi->td.mb.e_mbd.tile_ctx = &this_tile->tctx;
#if CONFIG_PVQ
        cpi->td.mb.pvq_q = &this_tile->pvq_q;
        cpi->td.mb.daala_enc.state.adapt = &this_tile->tctx.pvq_context;
#endif  // CONFIG_PVQ
#if !CONFIG_ANS
        aom_start_encode(&mode_bc, buf->data + data_offset);
        write_modes(cpi, &tile_info, &mode_bc, &tok, tok_end);
        assert(tok == tok_end);
        aom_stop_encode(&mode_bc);
        tile_size = mode_bc.pos;
#else
        buf_ans_write_init(buf_ans, buf->data + data_offset);
        write_modes(cpi, &tile_info, buf_ans, &tok, tok_end);
        assert(tok == tok_end);
        aom_buf_ans_flush(buf_ans);
        tile_size = buf_ans_write_end(buf_ans);
#endif  // !CONFIG_ANS
#if CONFIG_PVQ
        cpi->td.mb.pvq_q = NULL;
#endif
        buf->size = tile_size;

        // Record the maximum tile size we see, so we can compact headers later.
        *max_tile_size = AOMMAX(*max_tile_size, tile_size);

        if (have_tiles) {
          // tile header: size of this tile, or copy offset
          uint32_t tile_header = tile_size;
          const int tile_copy_mode =
              ((AOMMAX(cm->tile_width, cm->tile_height) << MI_SIZE_LOG2) <= 256)
                  ? 1
                  : 0;

          // If tile_copy_mode = 1, check if this tile is a copy tile.
          // Very low chances to have copy tiles on the key frames, so don't
          // search on key frames to reduce unnecessary search.
          if (cm->frame_type != KEY_FRAME && tile_copy_mode) {
            const int idendical_tile_offset =
                find_identical_tile(tile_row, tile_col, tile_buffers);

            if (idendical_tile_offset > 0) {
              tile_size = 0;
              tile_header = idendical_tile_offset | 0x80;
              tile_header <<= 24;
            }
          }

          mem_put_le32(buf->data, tile_header);
        }

        total_size += tile_size;
      }

      if (!is_last_col) {
        uint32_t col_size = total_size - col_offset - 4;
        mem_put_le32(dst + col_offset, col_size);

        // If it is not final packing, record the maximum tile column size we
        // see, otherwise, check if the tile size is out of the range.
        *max_tile_col_size = AOMMAX(*max_tile_col_size, col_size);
      }
    }
  } else {
#endif  // CONFIG_EXT_TILE
    write_uncompressed_header(cpi, &wb);

#if CONFIG_EXT_REFS
    if (cm->show_existing_frame) {
      total_size = aom_wb_bytes_written(&wb);
      return (uint32_t)total_size;
    }
#endif  // CONFIG_EXT_REFS

    // Write the tile length code
    tile_size_bytes_wb = wb;
    aom_wb_write_literal(&wb, 3, 2);

    /* Write a placeholder for the number of tiles in each tile group */
    tg_params_wb = wb;
    saved_offset = wb.bit_offset;
    if (have_tiles) {
      aom_wb_overwrite_literal(&wb, 3, n_log2_tiles);
      aom_wb_overwrite_literal(&wb, (1 << n_log2_tiles) - 1, n_log2_tiles);
    }

    /* Write a placeholder for the compressed header length */
    comp_hdr_len_wb = wb;
    aom_wb_write_literal(&wb, 0, 16);

    uncompressed_hdr_size = aom_wb_bytes_written(&wb);
    comp_hdr_size = write_compressed_header(cpi, dst + uncompressed_hdr_size);
    aom_wb_overwrite_literal(&comp_hdr_len_wb, (int)(comp_hdr_size), 16);
    hdr_size = uncompressed_hdr_size + comp_hdr_size;
    total_size += hdr_size;

    for (tile_row = 0; tile_row < tile_rows; tile_row++) {
      TileInfo tile_info;
      const int is_last_row = (tile_row == tile_rows - 1);
      av1_tile_set_row(&tile_info, cm, tile_row);

      for (tile_col = 0; tile_col < tile_cols; tile_col++) {
        const int tile_idx = tile_row * tile_cols + tile_col;
        TileBufferEnc *const buf = &tile_buffers[tile_row][tile_col];
        TileDataEnc *this_tile = &cpi->tile_data[tile_idx];
        const TOKENEXTRA *tok = tok_buffers[tile_row][tile_col];
        const TOKENEXTRA *tok_end = tok + cpi->tok_count[tile_row][tile_col];
        const int is_last_col = (tile_col == tile_cols - 1);
        const int is_last_tile = is_last_col && is_last_row;

        if ((!mtu_size && tile_count > tg_size) ||
            (mtu_size && tile_count && curr_tg_data_size >= mtu_size)) {
          // New tile group
          tg_count++;
          // We've exceeded the packet size
          if (tile_count > 1) {
            /* The last tile exceeded the packet size. The tile group size
               should therefore be tile_count-1.
               Move the last tile and insert headers before it
             */
            uint32_t old_total_size = total_size - tile_size - 4;
            memmove(dst + old_total_size + hdr_size, dst + old_total_size,
                    (tile_size + 4) * sizeof(uint8_t));
            // Copy uncompressed header
            memmove(dst + old_total_size, dst,
                    uncompressed_hdr_size * sizeof(uint8_t));
            // Write the number of tiles in the group into the last uncompressed
            // header before the one we've just inserted
            aom_wb_overwrite_literal(&tg_params_wb, tile_idx - tile_count,
                                     n_log2_tiles);
            aom_wb_overwrite_literal(&tg_params_wb, tile_count - 2,
                                     n_log2_tiles);
            // Update the pointer to the last TG params
            tg_params_wb.bit_offset = saved_offset + 8 * old_total_size;
            // Copy compressed header
            memmove(dst + old_total_size + uncompressed_hdr_size,
                    dst + uncompressed_hdr_size,
                    comp_hdr_size * sizeof(uint8_t));
            total_size += hdr_size;
            tile_count = 1;
            curr_tg_data_size = hdr_size + tile_size + 4;
          } else {
            // We exceeded the packet size in just one tile
            // Copy uncompressed header
            memmove(dst + total_size, dst,
                    uncompressed_hdr_size * sizeof(uint8_t));
            // Write the number of tiles in the group into the last uncompressed
            // header
            aom_wb_overwrite_literal(&tg_params_wb, tile_idx - tile_count,
                                     n_log2_tiles);
            aom_wb_overwrite_literal(&tg_params_wb, tile_count - 1,
                                     n_log2_tiles);
            tg_params_wb.bit_offset = saved_offset + 8 * total_size;
            // Copy compressed header
            memmove(dst + total_size + uncompressed_hdr_size,
                    dst + uncompressed_hdr_size,
                    comp_hdr_size * sizeof(uint8_t));
            total_size += hdr_size;
            tile_count = 0;
            curr_tg_data_size = hdr_size;
          }
        }
        tile_count++;
        av1_tile_set_col(&tile_info, cm, tile_col);

#if CONFIG_DEPENDENT_HORZTILES
        av1_tile_set_tg_boundary(&tile_info, cm, tile_row, tile_col);
#endif
        buf->data = dst + total_size;

        // The last tile does not have a header.
        if (!is_last_tile) total_size += 4;

        // Initialise tile context from the frame context
        this_tile->tctx = *cm->fc;
        cpi->td.mb.e_mbd.tile_ctx = &this_tile->tctx;
#if CONFIG_PVQ
        cpi->td.mb.pvq_q = &this_tile->pvq_q;
        cpi->td.mb.daala_enc.state.adapt = &this_tile->tctx.pvq_context;
#endif  // CONFIG_PVQ
#if CONFIG_ANS
        buf_ans_write_init(buf_ans, dst + total_size);
        write_modes(cpi, &tile_info, buf_ans, &tok, tok_end);
        assert(tok == tok_end);
        aom_buf_ans_flush(buf_ans);
        tile_size = buf_ans_write_end(buf_ans);
#else
      aom_start_encode(&mode_bc, dst + total_size);
      write_modes(cpi, &tile_info, &mode_bc, &tok, tok_end);
#if !CONFIG_LV_MAP
#if !CONFIG_PVQ
      assert(tok == tok_end);
#endif  // !CONFIG_PVQ
#endif  // !CONFIG_LV_MAP
      aom_stop_encode(&mode_bc);
      tile_size = mode_bc.pos;
#endif  // CONFIG_ANS
#if CONFIG_PVQ
        cpi->td.mb.pvq_q = NULL;
#endif

        assert(tile_size > 0);

        curr_tg_data_size += tile_size + 4;
        buf->size = tile_size;

        if (!is_last_tile) {
          *max_tile_size = AOMMAX(*max_tile_size, tile_size);
          // size of this tile
          mem_put_le32(buf->data, tile_size);
        }

        total_size += tile_size;
      }
    }
    // Write the final tile group size
    if (n_log2_tiles) {
      aom_wb_overwrite_literal(&tg_params_wb, (1 << n_log2_tiles) - tile_count,
                               n_log2_tiles);
      aom_wb_overwrite_literal(&tg_params_wb, tile_count - 1, n_log2_tiles);
    }
    // Remux if possible. TODO (Thomas Davies): do this for more than one tile
    // group
    if (have_tiles && tg_count == 1) {
      int data_size = total_size - (uncompressed_hdr_size + comp_hdr_size);
      data_size = remux_tiles(cm, dst + uncompressed_hdr_size + comp_hdr_size,
                              data_size, *max_tile_size, *max_tile_col_size,
                              &tile_size_bytes, &tile_col_size_bytes);
      total_size = data_size + uncompressed_hdr_size + comp_hdr_size;
      aom_wb_overwrite_literal(&tile_size_bytes_wb, tile_size_bytes - 1, 2);
    }

#if CONFIG_EXT_TILE
  }
#endif  // CONFIG_EXT_TILE
  return (uint32_t)total_size;
}

static void write_render_size(const AV1_COMMON *cm,
                              struct aom_write_bit_buffer *wb) {
  const int scaling_active = !av1_resize_unscaled(cm);
  aom_wb_write_bit(wb, scaling_active);
  if (scaling_active) {
    aom_wb_write_literal(wb, cm->render_width - 1, 16);
    aom_wb_write_literal(wb, cm->render_height - 1, 16);
  }
}

#if CONFIG_FRAME_SUPERRES
static void write_superres_scale(const AV1_COMMON *const cm,
                                 struct aom_write_bit_buffer *wb) {
  // First bit is whether to to scale or not
  if (cm->superres_scale_numerator == SCALE_DENOMINATOR) {
    aom_wb_write_bit(wb, 0);  // no scaling
  } else {
    aom_wb_write_bit(wb, 1);  // scaling, write scale factor
    aom_wb_write_literal(
        wb, cm->superres_scale_numerator - SUPERRES_SCALE_NUMERATOR_MIN,
        SUPERRES_SCALE_BITS);
  }
}
#endif  // CONFIG_FRAME_SUPERRES

static void write_frame_size(const AV1_COMMON *cm,
                             struct aom_write_bit_buffer *wb) {
#if CONFIG_FRAME_SUPERRES
  aom_wb_write_literal(wb, cm->superres_upscaled_width - 1, 16);
  aom_wb_write_literal(wb, cm->superres_upscaled_height - 1, 16);
  write_superres_scale(cm, wb);
#else
  aom_wb_write_literal(wb, cm->width - 1, 16);
  aom_wb_write_literal(wb, cm->height - 1, 16);
#endif  // CONFIG_FRAME_SUPERRES
  write_render_size(cm, wb);
}

static void write_frame_size_with_refs(AV1_COMP *cpi,
                                       struct aom_write_bit_buffer *wb) {
  AV1_COMMON *const cm = &cpi->common;
  int found = 0;

  MV_REFERENCE_FRAME ref_frame;
  for (ref_frame = LAST_FRAME; ref_frame <= ALTREF_FRAME; ++ref_frame) {
    YV12_BUFFER_CONFIG *cfg = get_ref_frame_buffer(cpi, ref_frame);

    if (cfg != NULL) {
#if CONFIG_FRAME_SUPERRES
      found = cm->superres_upscaled_width == cfg->y_crop_width &&
              cm->superres_upscaled_height == cfg->y_crop_height;
#else
      found =
          cm->width == cfg->y_crop_width && cm->height == cfg->y_crop_height;
#endif
      found &= cm->render_width == cfg->render_width &&
               cm->render_height == cfg->render_height;
    }
    aom_wb_write_bit(wb, found);
    if (found) {
#if CONFIG_FRAME_SUPERRES
      write_superres_scale(cm, wb);
#endif  // CONFIG_FRAME_SUPERRES
      break;
    }
  }

  if (!found) write_frame_size(cm, wb);
}

static void write_sync_code(struct aom_write_bit_buffer *wb) {
  aom_wb_write_literal(wb, AV1_SYNC_CODE_0, 8);
  aom_wb_write_literal(wb, AV1_SYNC_CODE_1, 8);
  aom_wb_write_literal(wb, AV1_SYNC_CODE_2, 8);
}

static void write_profile(BITSTREAM_PROFILE profile,
                          struct aom_write_bit_buffer *wb) {
  switch (profile) {
    case PROFILE_0: aom_wb_write_literal(wb, 0, 2); break;
    case PROFILE_1: aom_wb_write_literal(wb, 2, 2); break;
    case PROFILE_2: aom_wb_write_literal(wb, 1, 2); break;
    case PROFILE_3: aom_wb_write_literal(wb, 6, 3); break;
    default: assert(0);
  }
}

static void write_bitdepth_colorspace_sampling(
    AV1_COMMON *const cm, struct aom_write_bit_buffer *wb) {
  if (cm->profile >= PROFILE_2) {
    assert(cm->bit_depth > AOM_BITS_8);
    aom_wb_write_bit(wb, cm->bit_depth == AOM_BITS_10 ? 0 : 1);
  }
#if CONFIG_COLORSPACE_HEADERS
  aom_wb_write_literal(wb, cm->color_space, 5);
  aom_wb_write_literal(wb, cm->transfer_function, 5);
#else
  aom_wb_write_literal(wb, cm->color_space, 3);
#endif
  if (cm->color_space != AOM_CS_SRGB) {
    // 0: [16, 235] (i.e. xvYCC), 1: [0, 255]
    aom_wb_write_bit(wb, cm->color_range);
    if (cm->profile == PROFILE_1 || cm->profile == PROFILE_3) {
      assert(cm->subsampling_x != 1 || cm->subsampling_y != 1);
      aom_wb_write_bit(wb, cm->subsampling_x);
      aom_wb_write_bit(wb, cm->subsampling_y);
      aom_wb_write_bit(wb, 0);  // unused
    } else {
      assert(cm->subsampling_x == 1 && cm->subsampling_y == 1);
    }
#if CONFIG_COLORSPACE_HEADERS
    if (cm->subsampling_x == 1 && cm->subsampling_y == 1) {
      aom_wb_write_literal(wb, cm->chroma_sample_position, 2);
    }
#endif
  } else {
    assert(cm->profile == PROFILE_1 || cm->profile == PROFILE_3);
    aom_wb_write_bit(wb, 0);  // unused
  }
}

#if CONFIG_REFERENCE_BUFFER
void write_sequence_header(
#if CONFIG_EXT_TILE
    AV1_COMMON *const cm,
#endif  // CONFIG_EXT_TILE
    SequenceHeader *seq_params) {
  /* Placeholder for actually writing to the bitstream */
  seq_params->frame_id_numbers_present_flag =
#if CONFIG_EXT_TILE
      cm->large_scale_tile ? 0 :
#endif  // CONFIG_EXT_TILE
                           FRAME_ID_NUMBERS_PRESENT_FLAG;
  seq_params->frame_id_length_minus7 = FRAME_ID_LENGTH_MINUS7;
  seq_params->delta_frame_id_length_minus2 = DELTA_FRAME_ID_LENGTH_MINUS2;
}
#endif

#if CONFIG_EXT_INTER
static void write_compound_tools(const AV1_COMMON *cm,
                                 struct aom_write_bit_buffer *wb) {
  (void)cm;
  (void)wb;
#if CONFIG_INTERINTRA
  if (!frame_is_intra_only(cm) && cm->reference_mode != COMPOUND_REFERENCE) {
    aom_wb_write_bit(wb, cm->allow_interintra_compound);
  } else {
    assert(cm->allow_interintra_compound == 0);
  }
#endif  // CONFIG_INTERINTRA
#if CONFIG_WEDGE || CONFIG_COMPOUND_SEGMENT
#if CONFIG_COMPOUND_SINGLEREF
  if (!frame_is_intra_only(cm)) {
#else   // !CONFIG_COMPOUND_SINGLEREF
  if (!frame_is_intra_only(cm) && cm->reference_mode != SINGLE_REFERENCE) {
#endif  // CONFIG_COMPOUND_SINGLEREF
    aom_wb_write_bit(wb, cm->allow_masked_compound);
  } else {
    assert(cm->allow_masked_compound == 0);
  }
#endif  // CONFIG_WEDGE || CONFIG_COMPOUND_SEGMENT
}
#endif  // CONFIG_EXT_INTER

static void write_uncompressed_header(AV1_COMP *cpi,
                                      struct aom_write_bit_buffer *wb) {
  AV1_COMMON *const cm = &cpi->common;
  MACROBLOCKD *const xd = &cpi->td.mb.e_mbd;

#if CONFIG_REFERENCE_BUFFER
  /* TODO: Move outside frame loop or inside key-frame branch */
  write_sequence_header(
#if CONFIG_EXT_TILE
      cm,
#endif  // CONFIG_EXT_TILE
      &cpi->seq_params);
#endif

  aom_wb_write_literal(wb, AOM_FRAME_MARKER, 2);

  write_profile(cm->profile, wb);

#if CONFIG_EXT_TILE
  aom_wb_write_literal(wb, cm->large_scale_tile, 1);
#endif  // CONFIG_EXT_TILE

#if CONFIG_EXT_REFS
  // NOTE: By default all coded frames to be used as a reference
  cm->is_reference_frame = 1;

  if (cm->show_existing_frame) {
    RefCntBuffer *const frame_bufs = cm->buffer_pool->frame_bufs;
    const int frame_to_show = cm->ref_frame_map[cpi->existing_fb_idx_to_show];

    if (frame_to_show < 0 || frame_bufs[frame_to_show].ref_count < 1) {
      aom_internal_error(&cm->error, AOM_CODEC_UNSUP_BITSTREAM,
                         "Buffer %d does not contain a reconstructed frame",
                         frame_to_show);
    }
    ref_cnt_fb(frame_bufs, &cm->new_fb_idx, frame_to_show);

    aom_wb_write_bit(wb, 1);  // show_existing_frame
    aom_wb_write_literal(wb, cpi->existing_fb_idx_to_show, 3);

#if CONFIG_REFERENCE_BUFFER
    if (cpi->seq_params.frame_id_numbers_present_flag) {
      int frame_id_len = cpi->seq_params.frame_id_length_minus7 + 7;
      int display_frame_id = cm->ref_frame_id[cpi->existing_fb_idx_to_show];
      aom_wb_write_literal(wb, display_frame_id, frame_id_len);
      /* Add a zero byte to prevent emulation of superframe marker */
      /* Same logic as when when terminating the entropy coder */
      /* Consider to have this logic only one place */
      aom_wb_write_literal(wb, 0, 8);
    }
#endif

    return;
  } else {
#endif                        // CONFIG_EXT_REFS
    aom_wb_write_bit(wb, 0);  // show_existing_frame
#if CONFIG_EXT_REFS
  }
#endif  // CONFIG_EXT_REFS

  aom_wb_write_bit(wb, cm->frame_type);
  aom_wb_write_bit(wb, cm->show_frame);
  aom_wb_write_bit(wb, cm->error_resilient_mode);

#if CONFIG_REFERENCE_BUFFER
  cm->invalid_delta_frame_id_minus1 = 0;
  if (cpi->seq_params.frame_id_numbers_present_flag) {
    int frame_id_len = cpi->seq_params.frame_id_length_minus7 + 7;
    aom_wb_write_literal(wb, cm->current_frame_id, frame_id_len);
  }
#endif

  if (cm->frame_type == KEY_FRAME) {
    write_sync_code(wb);
    write_bitdepth_colorspace_sampling(cm, wb);
    write_frame_size(cm, wb);
#if CONFIG_ANS && ANS_MAX_SYMBOLS
    assert(cpi->common.ans_window_size_log2 >= 8);
    assert(cpi->common.ans_window_size_log2 < 24);
    aom_wb_write_literal(wb, cpi->common.ans_window_size_log2 - 8, 4);
#endif  // CONFIG_ANS && ANS_MAX_SYMBOLS
#if CONFIG_PALETTE || CONFIG_INTRABC
    aom_wb_write_bit(wb, cm->allow_screen_content_tools);
#endif  // CONFIG_PALETTE || CONFIG_INTRABC
  } else {
    if (!cm->show_frame) aom_wb_write_bit(wb, cm->intra_only);
#if CONFIG_PALETTE || CONFIG_INTRABC
    if (cm->intra_only) aom_wb_write_bit(wb, cm->allow_screen_content_tools);
#endif  // CONFIG_PALETTE || CONFIG_INTRABC
    if (!cm->error_resilient_mode) {
      if (cm->intra_only) {
        aom_wb_write_bit(wb,
                         cm->reset_frame_context == RESET_FRAME_CONTEXT_ALL);
      } else {
        aom_wb_write_bit(wb,
                         cm->reset_frame_context != RESET_FRAME_CONTEXT_NONE);
        if (cm->reset_frame_context != RESET_FRAME_CONTEXT_NONE)
          aom_wb_write_bit(wb,
                           cm->reset_frame_context == RESET_FRAME_CONTEXT_ALL);
      }
    }

#if CONFIG_EXT_REFS
    cpi->refresh_frame_mask = get_refresh_mask(cpi);
#endif  // CONFIG_EXT_REFS

    if (cm->intra_only) {
      write_sync_code(wb);
      write_bitdepth_colorspace_sampling(cm, wb);

#if CONFIG_EXT_REFS
      aom_wb_write_literal(wb, cpi->refresh_frame_mask, REF_FRAMES);
#else
      aom_wb_write_literal(wb, get_refresh_mask(cpi), REF_FRAMES);
#endif  // CONFIG_EXT_REFS
      write_frame_size(cm, wb);

#if CONFIG_ANS && ANS_MAX_SYMBOLS
      assert(cpi->common.ans_window_size_log2 >= 8);
      assert(cpi->common.ans_window_size_log2 < 24);
      aom_wb_write_literal(wb, cpi->common.ans_window_size_log2 - 8, 4);
#endif  // CONFIG_ANS && ANS_MAX_SYMBOLS
    } else {
      MV_REFERENCE_FRAME ref_frame;

#if CONFIG_EXT_REFS
      aom_wb_write_literal(wb, cpi->refresh_frame_mask, REF_FRAMES);
#else
      aom_wb_write_literal(wb, get_refresh_mask(cpi), REF_FRAMES);
#endif  // CONFIG_EXT_REFS

#if CONFIG_EXT_REFS
      if (!cpi->refresh_frame_mask) {
        // NOTE: "cpi->refresh_frame_mask == 0" indicates that the coded frame
        //       will not be used as a reference
        cm->is_reference_frame = 0;
      }
#endif  // CONFIG_EXT_REFS

      for (ref_frame = LAST_FRAME; ref_frame <= ALTREF_FRAME; ++ref_frame) {
        assert(get_ref_frame_map_idx(cpi, ref_frame) != INVALID_IDX);
        aom_wb_write_literal(wb, get_ref_frame_map_idx(cpi, ref_frame),
                             REF_FRAMES_LOG2);
        aom_wb_write_bit(wb, cm->ref_frame_sign_bias[ref_frame]);
#if CONFIG_REFERENCE_BUFFER
        if (cpi->seq_params.frame_id_numbers_present_flag) {
          int i = get_ref_frame_map_idx(cpi, ref_frame);
          int frame_id_len = cpi->seq_params.frame_id_length_minus7 + 7;
          int diff_len = cpi->seq_params.delta_frame_id_length_minus2 + 2;
          int delta_frame_id_minus1 =
              ((cm->current_frame_id - cm->ref_frame_id[i] +
                (1 << frame_id_len)) %
               (1 << frame_id_len)) -
              1;
          if (delta_frame_id_minus1 < 0 ||
              delta_frame_id_minus1 >= (1 << diff_len))
            cm->invalid_delta_frame_id_minus1 = 1;
          aom_wb_write_literal(wb, delta_frame_id_minus1, diff_len);
        }
#endif
      }

#if CONFIG_FRAME_SIZE
      if (cm->error_resilient_mode == 0) {
        write_frame_size_with_refs(cpi, wb);
      } else {
        write_frame_size(cm, wb);
      }
#else
      write_frame_size_with_refs(cpi, wb);
#endif

      aom_wb_write_bit(wb, cm->allow_high_precision_mv);

      fix_interp_filter(cm, cpi->td.counts);
      write_frame_interp_filter(cm->interp_filter, wb);
#if CONFIG_TEMPMV_SIGNALING
      if (!cm->error_resilient_mode) {
        aom_wb_write_bit(wb, cm->use_prev_frame_mvs);
      }
#endif
    }
  }

#if CONFIG_REFERENCE_BUFFER
  cm->refresh_mask = cm->frame_type == KEY_FRAME ? 0xFF : get_refresh_mask(cpi);
#endif

  if (!cm->error_resilient_mode) {
    aom_wb_write_bit(
        wb, cm->refresh_frame_context == REFRESH_FRAME_CONTEXT_FORWARD);
  }

  aom_wb_write_literal(wb, cm->frame_context_idx, FRAME_CONTEXTS_LOG2);

  assert(cm->mib_size == mi_size_wide[cm->sb_size]);
  assert(cm->mib_size == 1 << cm->mib_size_log2);
#if CONFIG_EXT_PARTITION
  assert(cm->sb_size == BLOCK_128X128 || cm->sb_size == BLOCK_64X64);
  aom_wb_write_bit(wb, cm->sb_size == BLOCK_128X128 ? 1 : 0);
#else
  assert(cm->sb_size == BLOCK_64X64);
#endif  // CONFIG_EXT_PARTITION

  encode_loopfilter(cm, wb);
  encode_quantization(cm, wb);
  encode_segmentation(cm, xd, wb);
#if CONFIG_DELTA_Q
  {
    int i;
    struct segmentation *const seg = &cm->seg;
    int segment_quantizer_active = 0;
    for (i = 0; i < MAX_SEGMENTS; i++) {
      if (segfeature_active(seg, i, SEG_LVL_ALT_Q)) {
        segment_quantizer_active = 1;
      }
    }

    if (cm->delta_q_present_flag)
      assert(segment_quantizer_active == 0 && cm->base_qindex > 0);
    if (segment_quantizer_active == 0 && cm->base_qindex > 0) {
      aom_wb_write_bit(wb, cm->delta_q_present_flag);
      if (cm->delta_q_present_flag) {
        aom_wb_write_literal(wb, OD_ILOG_NZ(cm->delta_q_res) - 1, 2);
        xd->prev_qindex = cm->base_qindex;
#if CONFIG_EXT_DELTA_Q
        assert(seg->abs_delta == SEGMENT_DELTADATA);
        aom_wb_write_bit(wb, cm->delta_lf_present_flag);
        if (cm->delta_lf_present_flag) {
          aom_wb_write_literal(wb, OD_ILOG_NZ(cm->delta_lf_res) - 1, 2);
          xd->prev_delta_lf_from_base = 0;
        }
#endif  // CONFIG_EXT_DELTA_Q
      }
    }
  }
#endif
#if CONFIG_CDEF
  if (!cm->all_lossless) {
    encode_cdef(cm, wb);
  }
#endif
#if CONFIG_LOOP_RESTORATION
  encode_restoration_mode(cm, wb);
#endif  // CONFIG_LOOP_RESTORATION
  write_tx_mode(cm, &cm->tx_mode, wb);

  if (cpi->allow_comp_inter_inter) {
    const int use_hybrid_pred = cm->reference_mode == REFERENCE_MODE_SELECT;
#if !CONFIG_REF_ADAPT
    const int use_compound_pred = cm->reference_mode != SINGLE_REFERENCE;
#endif  // !CONFIG_REF_ADAPT

    aom_wb_write_bit(wb, use_hybrid_pred);
#if !CONFIG_REF_ADAPT
    if (!use_hybrid_pred) aom_wb_write_bit(wb, use_compound_pred);
#endif  // !CONFIG_REF_ADAPT
  }
#if CONFIG_EXT_INTER
  write_compound_tools(cm, wb);
#endif  // CONFIG_EXT_INTER

#if CONFIG_EXT_TX
  aom_wb_write_bit(wb, cm->reduced_tx_set_used);
#endif  // CONFIG_EXT_TX

  write_tile_info(cm, wb);
}

#if CONFIG_GLOBAL_MOTION
static void write_global_motion_params(WarpedMotionParams *params,
                                       WarpedMotionParams *ref_params,
                                       aom_writer *w, int allow_hp) {
  TransformationType type = params->wmtype;
  int trans_bits;
  int trans_prec_diff;
  aom_write_bit(w, type != IDENTITY);
  if (type != IDENTITY) aom_write_literal(w, type - 1, GLOBAL_TYPE_BITS);

  switch (type) {
    case HOMOGRAPHY:
    case HORTRAPEZOID:
    case VERTRAPEZOID:
      if (type != HORTRAPEZOID)
        aom_write_signed_primitive_refsubexpfin(
            w, GM_ROW3HOMO_MAX + 1, SUBEXPFIN_K,
            (ref_params->wmmat[6] >> GM_ROW3HOMO_PREC_DIFF),
            (params->wmmat[6] >> GM_ROW3HOMO_PREC_DIFF));
      if (type != VERTRAPEZOID)
        aom_write_signed_primitive_refsubexpfin(
            w, GM_ROW3HOMO_MAX + 1, SUBEXPFIN_K,
            (ref_params->wmmat[7] >> GM_ROW3HOMO_PREC_DIFF),
            (params->wmmat[7] >> GM_ROW3HOMO_PREC_DIFF));
    // fallthrough intended
    case AFFINE:
    case ROTZOOM:
      aom_write_signed_primitive_refsubexpfin(
          w, GM_ALPHA_MAX + 1, SUBEXPFIN_K,
          (ref_params->wmmat[2] >> GM_ALPHA_PREC_DIFF) -
              (1 << GM_ALPHA_PREC_BITS),
          (params->wmmat[2] >> GM_ALPHA_PREC_DIFF) - (1 << GM_ALPHA_PREC_BITS));
      if (type != VERTRAPEZOID)
        aom_write_signed_primitive_refsubexpfin(
            w, GM_ALPHA_MAX + 1, SUBEXPFIN_K,
            (ref_params->wmmat[3] >> GM_ALPHA_PREC_DIFF),
            (params->wmmat[3] >> GM_ALPHA_PREC_DIFF));
      if (type >= AFFINE) {
        if (type != HORTRAPEZOID)
          aom_write_signed_primitive_refsubexpfin(
              w, GM_ALPHA_MAX + 1, SUBEXPFIN_K,
              (ref_params->wmmat[4] >> GM_ALPHA_PREC_DIFF),
              (params->wmmat[4] >> GM_ALPHA_PREC_DIFF));
        aom_write_signed_primitive_refsubexpfin(
            w, GM_ALPHA_MAX + 1, SUBEXPFIN_K,
            (ref_params->wmmat[5] >> GM_ALPHA_PREC_DIFF) -
                (1 << GM_ALPHA_PREC_BITS),
            (params->wmmat[5] >> GM_ALPHA_PREC_DIFF) -
                (1 << GM_ALPHA_PREC_BITS));
      }
    // fallthrough intended
    case TRANSLATION:
      trans_bits = (type == TRANSLATION) ? GM_ABS_TRANS_ONLY_BITS - !allow_hp
                                         : GM_ABS_TRANS_BITS;
      trans_prec_diff = (type == TRANSLATION)
                            ? GM_TRANS_ONLY_PREC_DIFF + !allow_hp
                            : GM_TRANS_PREC_DIFF;
      aom_write_signed_primitive_refsubexpfin(
          w, (1 << trans_bits) + 1, SUBEXPFIN_K,
          (ref_params->wmmat[0] >> trans_prec_diff),
          (params->wmmat[0] >> trans_prec_diff));
      aom_write_signed_primitive_refsubexpfin(
          w, (1 << trans_bits) + 1, SUBEXPFIN_K,
          (ref_params->wmmat[1] >> trans_prec_diff),
          (params->wmmat[1] >> trans_prec_diff));
      break;
    case IDENTITY: break;
    default: assert(0);
  }
}

static void write_global_motion(AV1_COMP *cpi, aom_writer *w) {
  AV1_COMMON *const cm = &cpi->common;
  int frame;
  YV12_BUFFER_CONFIG *ref_buf;
  for (frame = LAST_FRAME; frame <= ALTREF_FRAME; ++frame) {
    ref_buf = get_ref_frame_buffer(cpi, frame);
    if (cpi->source->y_crop_width == ref_buf->y_crop_width &&
        cpi->source->y_crop_height == ref_buf->y_crop_height) {
      write_global_motion_params(&cm->global_motion[frame],
                                 &cm->prev_frame->global_motion[frame], w,
                                 cm->allow_high_precision_mv);
    } else {
      assert(cm->global_motion[frame].wmtype == IDENTITY &&
             "Invalid warp type for frames of different resolutions");
    }
    /*
    printf("Frame %d/%d: Enc Ref %d (used %d): %d %d %d %d\n",
           cm->current_video_frame, cm->show_frame, frame,
           cpi->global_motion_used[frame], cm->global_motion[frame].wmmat[0],
           cm->global_motion[frame].wmmat[1], cm->global_motion[frame].wmmat[2],
           cm->global_motion[frame].wmmat[3]);
           */
  }
}
#endif

static uint32_t write_compressed_header(AV1_COMP *cpi, uint8_t *data) {
  AV1_COMMON *const cm = &cpi->common;
#if CONFIG_SUPERTX
  MACROBLOCKD *const xd = &cpi->td.mb.e_mbd;
#endif  // CONFIG_SUPERTX
  FRAME_CONTEXT *const fc = cm->fc;
  aom_writer *header_bc;
  int i;
#if !CONFIG_NEW_MULTISYMBOL
  FRAME_COUNTS *counts = cpi->td.counts;
  int j;
#endif

  const int probwt = cm->num_tg;
  (void)probwt;
  (void)i;
  (void)fc;

#if CONFIG_ANS
  int header_size;
  header_bc = &cpi->buf_ans;
  buf_ans_write_init(header_bc, data);
#else
  aom_writer real_header_bc;
  header_bc = &real_header_bc;
  aom_start_encode(header_bc, data);
#endif

#if CONFIG_LOOP_RESTORATION
  encode_restoration(cm, header_bc);
#endif  // CONFIG_LOOP_RESTORATION
#if CONFIG_RECT_TX_EXT && (CONFIG_EXT_TX || CONFIG_VAR_TX)
  if (cm->tx_mode == TX_MODE_SELECT)
    av1_cond_prob_diff_update(header_bc, &cm->fc->quarter_tx_size_prob,
                              cm->counts.quarter_tx_size, probwt);
#endif
#if CONFIG_LV_MAP
  av1_write_txb_probs(cpi, header_bc);
#endif  // CONFIG_LV_MAP

#if CONFIG_VAR_TX && !CONFIG_NEW_MULTISYMBOL
  update_txfm_partition_probs(cm, header_bc, counts, probwt);
#endif

#if !CONFIG_NEW_MULTISYMBOL
  update_skip_probs(cm, header_bc, counts);
#endif

  if (frame_is_intra_only(cm)) {
    av1_copy(cm->fc->kf_y_cdf, av1_kf_y_mode_cdf);

#if CONFIG_INTRABC
    if (cm->allow_screen_content_tools) {
      av1_cond_prob_diff_update(header_bc, &fc->intrabc_prob,
                                cm->counts.intrabc, probwt);
    }
#endif
  } else {
#if !CONFIG_NEW_MULTISYMBOL
    update_inter_mode_probs(cm, header_bc, counts);
#endif
#if CONFIG_EXT_INTER
#if CONFIG_INTERINTRA
    if (cm->reference_mode != COMPOUND_REFERENCE &&
        cm->allow_interintra_compound) {
#if !CONFIG_NEW_MULTISYMBOL
      for (i = 0; i < BLOCK_SIZE_GROUPS; i++) {
        if (is_interintra_allowed_bsize_group(i)) {
          av1_cond_prob_diff_update(header_bc, &fc->interintra_prob[i],
                                    cm->counts.interintra[i], probwt);
        }
      }
#endif
#if CONFIG_WEDGE && !CONFIG_NEW_MULTISYMBOL
#if CONFIG_EXT_PARTITION_TYPES
      int block_sizes_to_update = BLOCK_SIZES_ALL;
#else
      int block_sizes_to_update = BLOCK_SIZES;
#endif
      for (i = 0; i < block_sizes_to_update; i++) {
        if (is_interintra_allowed_bsize(i) && is_interintra_wedge_used(i))
          av1_cond_prob_diff_update(header_bc, &fc->wedge_interintra_prob[i],
                                    cm->counts.wedge_interintra[i], probwt);
      }
#endif  // CONFIG_WEDGE && CONFIG_NEW_MULTISYMBOL
    }
#endif  // CONFIG_INTERINTRA
#endif  // CONFIG_EXT_INTER

#if CONFIG_MOTION_VAR || CONFIG_WARPED_MOTION
#if CONFIG_NCOBMC_ADAPT_WEIGHT
    for (i = ADAPT_OVERLAP_BLOCK_8X8; i < ADAPT_OVERLAP_BLOCKS; ++i) {
      prob_diff_update(av1_ncobmc_mode_tree, fc->ncobmc_mode_prob[i],
                       counts->ncobmc_mode[i], MAX_NCOBMC_MODES, probwt,
                       header_bc);
    }
#endif
#endif  // CONFIG_MOTION_VAR || CONFIG_WARPED_MOTION

#if !CONFIG_NEW_MULTISYMBOL
    for (i = 0; i < INTRA_INTER_CONTEXTS; i++)
      av1_cond_prob_diff_update(header_bc, &fc->intra_inter_prob[i],
                                counts->intra_inter[i], probwt);
#endif

#if !CONFIG_NEW_MULTISYMBOL
    if (cpi->allow_comp_inter_inter) {
      const int use_hybrid_pred = cm->reference_mode == REFERENCE_MODE_SELECT;
      if (use_hybrid_pred)
        for (i = 0; i < COMP_INTER_CONTEXTS; i++)
          av1_cond_prob_diff_update(header_bc, &fc->comp_inter_prob[i],
                                    counts->comp_inter[i], probwt);
    }

    if (cm->reference_mode != COMPOUND_REFERENCE) {
      for (i = 0; i < REF_CONTEXTS; i++) {
        for (j = 0; j < (SINGLE_REFS - 1); j++) {
          av1_cond_prob_diff_update(header_bc, &fc->single_ref_prob[i][j],
                                    counts->single_ref[i][j], probwt);
        }
      }
    }

    if (cm->reference_mode != SINGLE_REFERENCE) {
#if CONFIG_EXT_COMP_REFS
      for (i = 0; i < COMP_REF_TYPE_CONTEXTS; i++)
        av1_cond_prob_diff_update(header_bc, &fc->comp_ref_type_prob[i],
                                  counts->comp_ref_type[i], probwt);

      for (i = 0; i < UNI_COMP_REF_CONTEXTS; i++)
        for (j = 0; j < (UNIDIR_COMP_REFS - 1); j++)
          av1_cond_prob_diff_update(header_bc, &fc->uni_comp_ref_prob[i][j],
                                    counts->uni_comp_ref[i][j], probwt);
#endif  // CONFIG_EXT_COMP_REFS

      for (i = 0; i < REF_CONTEXTS; i++) {
#if CONFIG_EXT_REFS
        for (j = 0; j < (FWD_REFS - 1); j++) {
          av1_cond_prob_diff_update(header_bc, &fc->comp_ref_prob[i][j],
                                    counts->comp_ref[i][j], probwt);
        }
        for (j = 0; j < (BWD_REFS - 1); j++) {
          av1_cond_prob_diff_update(header_bc, &fc->comp_bwdref_prob[i][j],
                                    counts->comp_bwdref[i][j], probwt);
        }
#else
        for (j = 0; j < (COMP_REFS - 1); j++) {
          av1_cond_prob_diff_update(header_bc, &fc->comp_ref_prob[i][j],
                                    counts->comp_ref[i][j], probwt);
        }
#endif  // CONFIG_EXT_REFS
      }
    }
#endif  // CONFIG_NEW_MULTISYMBOL

#if CONFIG_EXT_INTER && CONFIG_COMPOUND_SINGLEREF
    for (i = 0; i < COMP_INTER_MODE_CONTEXTS; i++)
      av1_cond_prob_diff_update(header_bc, &fc->comp_inter_mode_prob[i],
                                counts->comp_inter_mode[i], probwt);
#endif  // CONFIG_EXT_INTER && CONFIG_COMPOUND_SINGLEREF

#if !CONFIG_NEW_MULTISYMBOL
    av1_write_nmv_probs(cm, cm->allow_high_precision_mv, header_bc, counts->mv);
#endif
#if CONFIG_SUPERTX
    if (!xd->lossless[0]) update_supertx_probs(cm, probwt, header_bc);
#endif  // CONFIG_SUPERTX
#if CONFIG_GLOBAL_MOTION
    write_global_motion(cpi, header_bc);
#endif  // CONFIG_GLOBAL_MOTION
  }
#if CONFIG_ANS
  aom_buf_ans_flush(header_bc);
  header_size = buf_ans_write_end(header_bc);
  assert(header_size <= 0xffff);
  return header_size;
#else
  aom_stop_encode(header_bc);
  assert(header_bc->pos <= 0xffff);
  return header_bc->pos;
#endif  // CONFIG_ANS
}

static int choose_size_bytes(uint32_t size, int spare_msbs) {
  // Choose the number of bytes required to represent size, without
  // using the 'spare_msbs' number of most significant bits.

  // Make sure we will fit in 4 bytes to start with..
  if (spare_msbs > 0 && size >> (32 - spare_msbs) != 0) return -1;

  // Normalise to 32 bits
  size <<= spare_msbs;

  if (size >> 24 != 0)
    return 4;
  else if (size >> 16 != 0)
    return 3;
  else if (size >> 8 != 0)
    return 2;
  else
    return 1;
}

static void mem_put_varsize(uint8_t *const dst, const int sz, const int val) {
  switch (sz) {
    case 1: dst[0] = (uint8_t)(val & 0xff); break;
    case 2: mem_put_le16(dst, val); break;
    case 3: mem_put_le24(dst, val); break;
    case 4: mem_put_le32(dst, val); break;
    default: assert(0 && "Invalid size"); break;
  }
}
static int remux_tiles(const AV1_COMMON *const cm, uint8_t *dst,
                       const uint32_t data_size, const uint32_t max_tile_size,
                       const uint32_t max_tile_col_size,
                       int *const tile_size_bytes,
                       int *const tile_col_size_bytes) {
  // Choose the tile size bytes (tsb) and tile column size bytes (tcsb)
  int tsb;
  int tcsb;

#if CONFIG_EXT_TILE
  if (cm->large_scale_tile) {
    // The top bit in the tile size field indicates tile copy mode, so we
    // have 1 less bit to code the tile size
    tsb = choose_size_bytes(max_tile_size, 1);
    tcsb = choose_size_bytes(max_tile_col_size, 0);
  } else {
#endif  // CONFIG_EXT_TILE
    tsb = choose_size_bytes(max_tile_size, 0);
    tcsb = 4;  // This is ignored
    (void)max_tile_col_size;
#if CONFIG_EXT_TILE
  }
#endif  // CONFIG_EXT_TILE

  assert(tsb > 0);
  assert(tcsb > 0);

  *tile_size_bytes = tsb;
  *tile_col_size_bytes = tcsb;

  if (tsb == 4 && tcsb == 4) {
    return data_size;
  } else {
    uint32_t wpos = 0;
    uint32_t rpos = 0;

#if CONFIG_EXT_TILE
    if (cm->large_scale_tile) {
      int tile_row;
      int tile_col;

      for (tile_col = 0; tile_col < cm->tile_cols; tile_col++) {
        // All but the last column has a column header
        if (tile_col < cm->tile_cols - 1) {
          uint32_t tile_col_size = mem_get_le32(dst + rpos);
          rpos += 4;

          // Adjust the tile column size by the number of bytes removed
          // from the tile size fields.
          tile_col_size -= (4 - tsb) * cm->tile_rows;

          mem_put_varsize(dst + wpos, tcsb, tile_col_size);
          wpos += tcsb;
        }

        for (tile_row = 0; tile_row < cm->tile_rows; tile_row++) {
          // All, including the last row has a header
          uint32_t tile_header = mem_get_le32(dst + rpos);
          rpos += 4;

          // If this is a copy tile, we need to shift the MSB to the
          // top bit of the new width, and there is no data to copy.
          if (tile_header >> 31 != 0) {
            if (tsb < 4) tile_header >>= 32 - 8 * tsb;
            mem_put_varsize(dst + wpos, tsb, tile_header);
            wpos += tsb;
          } else {
            mem_put_varsize(dst + wpos, tsb, tile_header);
            wpos += tsb;

            memmove(dst + wpos, dst + rpos, tile_header);
            rpos += tile_header;
            wpos += tile_header;
          }
        }
      }
    } else {
#endif  // CONFIG_EXT_TILE
      const int n_tiles = cm->tile_cols * cm->tile_rows;
      int n;

      for (n = 0; n < n_tiles; n++) {
        int tile_size;

        if (n == n_tiles - 1) {
          tile_size = data_size - rpos;
        } else {
          tile_size = mem_get_le32(dst + rpos);
          rpos += 4;
          mem_put_varsize(dst + wpos, tsb, tile_size);
          wpos += tsb;
        }

        memmove(dst + wpos, dst + rpos, tile_size);

        rpos += tile_size;
        wpos += tile_size;
      }
#if CONFIG_EXT_TILE
    }
#endif  // CONFIG_EXT_TILE

    assert(rpos > wpos);
    assert(rpos == data_size);

    return wpos;
  }
}

void av1_pack_bitstream(AV1_COMP *const cpi, uint8_t *dst, size_t *size) {
  uint8_t *data = dst;
  uint32_t data_size;
#if CONFIG_EXT_TILE
  AV1_COMMON *const cm = &cpi->common;
  uint32_t compressed_header_size = 0;
  uint32_t uncompressed_header_size;
  struct aom_write_bit_buffer saved_wb;
  struct aom_write_bit_buffer wb = { data, 0 };
  const int have_tiles = cm->tile_cols * cm->tile_rows > 1;
  int tile_size_bytes;
  int tile_col_size_bytes;
#endif  // CONFIG_EXT_TILE
  unsigned int max_tile_size;
  unsigned int max_tile_col_size;

#if CONFIG_BITSTREAM_DEBUG
  bitstream_queue_reset_write();
#endif

#if CONFIG_EXT_TILE
  if (cm->large_scale_tile) {
    // Write the uncompressed header
    write_uncompressed_header(cpi, &wb);

#if CONFIG_EXT_REFS
    if (cm->show_existing_frame) {
      *size = aom_wb_bytes_written(&wb);
      return;
    }
#endif  // CONFIG_EXT_REFS

    // We do not know these in advance. Output placeholder bit.
    saved_wb = wb;
    // Write tile size magnitudes
    if (have_tiles) {
      // Note that the last item in the uncompressed header is the data
      // describing tile configuration.
      // Number of bytes in tile column size - 1
      aom_wb_write_literal(&wb, 0, 2);

      // Number of bytes in tile size - 1
      aom_wb_write_literal(&wb, 0, 2);
    }
    // Size of compressed header
    aom_wb_write_literal(&wb, 0, 16);

    uncompressed_header_size = (uint32_t)aom_wb_bytes_written(&wb);
    data += uncompressed_header_size;

    aom_clear_system_state();

    // Write the compressed header
    compressed_header_size = write_compressed_header(cpi, data);
    data += compressed_header_size;

    // Write the encoded tile data
    data_size = write_tiles(cpi, data, &max_tile_size, &max_tile_col_size);
  } else {
#endif  // CONFIG_EXT_TILE
    data_size = write_tiles(cpi, data, &max_tile_size, &max_tile_col_size);
#if CONFIG_EXT_TILE
  }
#endif  // CONFIG_EXT_TILE
#if CONFIG_EXT_TILE
  if (cm->large_scale_tile) {
    if (have_tiles) {
      data_size =
          remux_tiles(cm, data, data_size, max_tile_size, max_tile_col_size,
                      &tile_size_bytes, &tile_col_size_bytes);
    }

    data += data_size;

    // Now fill in the gaps in the uncompressed header.
    if (have_tiles) {
      assert(tile_col_size_bytes >= 1 && tile_col_size_bytes <= 4);
      aom_wb_write_literal(&saved_wb, tile_col_size_bytes - 1, 2);

      assert(tile_size_bytes >= 1 && tile_size_bytes <= 4);
      aom_wb_write_literal(&saved_wb, tile_size_bytes - 1, 2);
    }
    // TODO(jbb): Figure out what to do if compressed_header_size > 16 bits.
    assert(compressed_header_size <= 0xffff);
    aom_wb_write_literal(&saved_wb, compressed_header_size, 16);
  } else {
#endif  // CONFIG_EXT_TILE
    data += data_size;
#if CONFIG_EXT_TILE
  }
#endif  // CONFIG_EXT_TILE
#if CONFIG_ANS && ANS_REVERSE
  // Avoid aliasing the superframe index
  *data++ = 0;
#endif
  *size = data - dst;
}